The Oxford Handbook of Negation 9780198830528, 0198830521

Citation preview

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   

NEGATION

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

O X F O R D H A N D B O O K S IN LI N G U I S T I C S Recently published

THE OXFORD HANDBOOK OF POLYSYNTHESIS Edited by Michael Fortescue, Marianne Mithun, and Nicholas Evans

THE OXFORD HANDBOOK OF EVIDENTIALITY Edited by Alexandra Y. Aikhenvald

THE OXFORD HANDBOOK OF LANGUAGE POLICY AND PLANNING Edited by James W. Tollefson and Miguel Pérez-Milans

THE OXFORD HANDBOOK OF PERSIAN LINGUISTICS Edited by Anousha Sedighi and Pouneh Shabani-Jadidi

THE OXFORD HANDBOOK OF ENDANGERED LANGUAGES Edited by Kenneth L. Rehg and Lyle Campbell

THE OXFORD HANDBOOK OF ELLIPSIS Edited by Jeroen van Craenenbroeck and Tanja Temmerman

THE OXFORD HANDBOOK OF LYING Edited by Jörg Meibauer

THE OXFORD HANDBOOK OF TABOO WORDS AND LANGUAGE Edited by Keith Allan

THE OXFORD HANDBOOK OF MORPHOLOGICAL THEORY Edited by Jenny Audring and Francesca Masini

THE OXFORD HANDBOOK OF REFERENCE Edited by Jeanette Gundel and Barbara Abbott

THE OXFORD HANDBOOK OF EXPERIMENTAL SEMANTICS AND PRAGMATICS Edited by Chris Cummins and Napoleon Katsos

THE OXFORD HANDBOOK OF EVENT STRUCTURE Edited by Robert Truswell

THE OXFORD HANDBOOK OF LANGUAGE ATTRITION Edited by Monika S. Schmid and Barbara Köpke

THE OXFORD HANDBOOK OF LANGUAGE CONTACT Edited by Anthony P. Grant

THE OXFORD HANDBOOK OF ENGLISH GRAMMAR Edited by Bas Aarts, Jill Bowie, and Gergana Popova

THE OXFORD HANDBOOK OF AFRICAN LANGUAGES Edited by Rainer Vossen and Gerrit J. Dimmendaal

THE OXFORD HANDBOOK OF NEGATION Edited by Viviane Déprez and M.Teresa Espinal For a complete list of Oxford Handbooks in Linguistics please see pp. –.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   

.........................................................................................................................................

NEGATION ......................................................................................................................................... Edited by

VIVIANE DÉPREZ and

M.TERESA ESPINAL

1

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

3

Great Clarendon Street, Oxford,  , United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries © editorial matter and organization Viviane Déprez and M.Teresa Espinal  © the chapters their several authors  The moral rights of the authors have been asserted First Edition published in  Impression:  All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press  Madison Avenue, New York, NY , United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number:  ISBN –––– Printed and bound by CPI Group (UK) Ltd, Croydon,   Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

C................................  x xi xix

Acknowledgments

Abbreviations The Contributors . Introduction: Negation in language and beyond V D´  M.T E



PART I FUNDAMENTALS . Negation and opposition: Contradiction and contrariety in logic and language L R. H



. Negative predicates: Incorporated negation J M



. Denial D R



. Types of negation K D C



. Affixal negation S J



PART II QUESTIONS IN THE SYNTAX OF NEGATION . The typology of negation J   A  O K



. The morpho-syntactic nature of the negative marker C G



OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

vi



. The possible positioning of negation C P



. Negation and constituent ordering: Case studies E P



. The expression of negation in sign languages J Q



PART III NEGATION AT THE SYNTAX-SEMANTICS INTERFACE . Neg-raising L R. H



. Intervention effects with negation C M



. Form and function of negative, tag, and rhetorical questions M R



. Expletive negation D D



. Calculating the scope of negation: Interaction of negation with quantifiers N F



PART IV SEMANTICS AND PRAGMATICS OF NEGATION . Modals and negation N F  S I



. Negation in event semantics B S



. Negation and alternatives: Interaction with focus constituents A F˘ ˘ ¸



OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



vii

. Metalinguistic negation A M M



. Negation and presupposition D B  K D



PART V NEGATIVE DEPENDENCIES . Negative polarity items L M. T



. Minimizers and maximizers as different types of polarity items S T



. Negative quantifiers H Z



. Negative fragment answers A W



. Negative concord and the nature of negative concord items A G



. Double negation readings H̈  S



PART VI SYNCHRONIC AND DIACHRONIC VARIATION IN NEGATION . Quantitative studies of the use of negative (dependent) expressions P W



. Negation in non-standard varieties C T  F B



. The negative cycle and beyond A B



. Evolution of negative dependencies C G



OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

viii



. The role of pragmatics in negation change P Lˊ



PART VII EMERGENCE AND ACQUISITION OF NEGATION . Evolutionary precursors of negation in non-human reasoning M B, J C,  C J. Vö



. Cognitive precursors of negation in preverbal infants J-Rˊ H



. Negation and first language acquisition R T



. Negation in L acquisition and beyond L Sˊ  J A



PART VIII EXPERIMENTAL INVESTIGATIONS OF NEGATION . Understanding negation: Issues in the processing of negation B K  C D



. Negative polarity illusions H M  C P



. Negation, prosody, and gesture P P  M.T E



. Negation and the brain: Experiments in health and in focal brain disease, and their theoretical implications  Y G, V J, I D, Mˊ  E Sˊ, Mˊ  F, P P, Y L,  K A . Individual differences in processing of negative operators: Implications for bilinguals V D. D



OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



ix

. The neurology of negation: fMRI, ERP, and aphasia K Rø C



. The neurobiology of lexical and sentential negation L P  M  V



References Index

 

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

A

.....................................................................

We gratefully acknowledge the work of all our colleagues who have contributed a chapter to this volume, as well as Oxford University Press for having accepted our proposal. We would also like to deeply thank all the internal and external reviewers who have generously given their time and greatly contributed to improving the quality of this handbook, sometimes at very short notice. For their invaluable help, we are extremely grateful to: Maria Barouni, Heather Burnett, Paola Cepeda, Lucas Champollion, Chris Collins, Annabel Cormack, Urtzi Etxeberria, Ricardo Etxepare, Manuel García Carpintero, Anna Gavarró, Ely van Gelderen, Raquel González, Julia Herschensohn, Jack Hoeksema, Michael Israel, Rafal Jónczyk, Marie Labelle, Josep Macià, Alda Mari, Gary Marr, Eric Mathieu, Matti Miestamo, Aliyah Morgenstern, Vincenzo Moscati, Ronice Müller de Quadros, Ann Nordmeyer, Marloes Oomen, Ignacio Palacios Martínez, Doris Penka, Shana Poplack, Paul P. Portner, Ljiljana Progovac, Cristina Sánchez López, Maria Spychalska, Lilian Teixeira de Sousa, Lyn Tieu, Ljuba Veselinova, Xavier Villalba, and Aydogan Yamilaz. Josep Hornos (UAB) also deserves a special mention and our gratitude for his continuous technical support. We thank Evripidis Tsiakmakis (UAB) for his editorial assistance. Finally, we would like this volume to serve as a tribute to Larry Horn’s incomparable contribution to the study of negation. Not only did Larry graciously contribute two chapters to this handbook, but the extensive reference to his work that is visible in almost every chapter speaks volumes to the importance of his influence. Just as its presence pervades these pages, we are certain that his opus will continue to inspire generations of ‘negaholics’, present and future, as much as it has addicted us.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

A  ......................................................  SG/-sg.  h/hr/-hr/hrs m P SG  > -sg. A0 A AAPCAppE ABL ABS ACC ACM ACT Adv AdvP AFF AFFIRM AGR AgrP ALT API AQ ASL Asp ASS Auslan AUX BA BelAd BelSp BNC BOLD

st person first person singular nd person two hours two minutes second person second person singular rd person rd person agent þ rd person patient rd person singular (A-bar) non-argumental position nd person active (subject agreement) Audio-Aligned and Parsed Corpus of Appalachian English Ablative absolutive accusative Association for Computing Machinery active adverb Adverb Phrase affect affirmative agreement Agreement Phrase set of alternatives Affective Polarity Item Autism-spectrum Quotient American Sign Language Aspect assertive Australian Sign Language Auxiliary Brodmann Area addressee’s belief speaker’s belief British National Corpus Blood Oxygen Level Dependent

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

xii



BSL C C0 CE CG CGN CGSpþAD CI CL CI CLC CLF CNG COMPL CONN Convx(w’) COP COR CP CSL CSLI CSP CU DAT DCAd DCSp DE DEC DECL DEM DEO DEP DescN DET df DGS DISJ diss DIST DN DOHP DP DUR e

British Sign Language complementizer head of complementizer phrase Contextual Evidence Bias Common Ground corpus gesproken Nederlands Common Ground shared by both speaker and addressee conventional implicature Clitic ‘ci’ complementizer Cornell Linguistics Circle classifier connegative completive connective set of worlds where all the conversational goals of x at w’ are fulfilled copula corrective clause Complementizer Phrase Chinese Sign Language Center for the Study of Language and Information Cortical Silent Period Cloud of Unknowing dative Discourse Commitments of the addressee Discourse Commitments of the speaker downward entailment, downward entailing DE complexity declarative demonstrative deontic dependent descriptive Negation determiner definition German Sign Language disjunction dissertation distal double negation Dante Oral History Project determiner Phrase durative event

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 E EA EEG ELIM EMG EMPH EN EP EPI Epix(w) EPP ER ERD ERG ERP ERS ETL EVID EVAL Excl EXCLU F FCI FFLV FI Fin FinP fMRI FOC Focº FocP ForceP FORM FUT GER GESL GSL h HAB H-clause HKSL HiNQ HOD I

Even-exhaustification Egyptian Arabic Electroencephalography elimination electromyography emphatic expletive negation European Portuguese epistemic Set of epistemically accessible worlds of x at w Extended Projection Principle Error Rate Event-Related Desynchronization ergative Event-Related Potential Event-Related Synchronization Extended Term Logic evidential evaluative Exclamative Exclusive feminine Free Choice Item fixed form of the lexical verb final suffix Finiteness Finiteness Phrase functional magnetic resonance imaging focus Focus head Focus Phrase Force Phrase formative verb prefix future gerund Georgian Sign Language Greek Sign Language history habitual Horn-clause Hong Kong Sign Language High Negation Question hodiernal informative (Chapter )

xiii

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

xiv



I IMM IMP INCL IND [iNEG] INF INT IP IPFV IRR IrSL ISQ IsSL JBA JC L L LEM LF Libras LIL LIS LIU LLC LNC LOB LOC LowNQ LP LSC LSQ m M MAL Max-contrary MCP ME MEG MEP MIT MN MOD MRI

Inflection (Chapter ) immediate imperative inclusive indicative interpretable negative formal feature infinitive Interrogative Inflection Phrase imperfective irrealis Irish Sign Language Information Seeking Question Israeli Sign Language Jewish Babylonian Aramaic Jespersen’s Cycle First language Second language Law of Excluded Middle Logical Form Brazilian Sign Language Lebanese Sign Language Italian Sign Language Jordanian Sign Language London-Lund Corpus of Present-day English Law of Non-Contradiction Lancaster-Oslo-Bergen corpus locative Low Negation Question lexical presupposition Catalan Sign Language Quebec Sign Language moment masculine malefactive maximum contrary Main Clause Phenomena Middle English (– ) Magnetoencephalography motor-evoked potential Massachusetts Institute of Technology Metalinguistic Negation modal Magnetic Resonance Imaging

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 MRKR ms N N NAI NC NCI NEC NEG Neg Neg NegDP NEGEX NegP NEG/POS NEG-Q Neg-Raising NEG-V NF-comp NGT NI NM NMLZ NOM NONIND NP NPFV NPI NVN NZSL O OB OBL OT P P PART PASS PAST p-cancelling PDE PERM PERS PERSIS

marker (meaning not specified or not relevant) millisecond Noun Negative  Negative Auxiliary Inversion negative concord Negative Concord Item necessity negation, negative preverbal negative marker postverbal negative marker Negative Determiner Phrase negative existential Negation Phrase Negative/Positive negation above quantifier negation raising negation above verb / below quantifier non-factive complementizer Sign Language of the Netherlands negative indefinite negative marker nominalizer nominative non-indicative noun phrase non-perfective Negative Polarity Item Noun Verb Noun New Zealand Sign Language Only-exhaustification Original Speaker Bias oblique Optimality Theory Preposition Positive  partitive passive past tense negation presupposition cancelling negation present-day English permissive personal persistive

xv

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

xvi



PF PI PL PNP POL Pol-Int PolP POS POSQ POSS POSS PP PP PPCEME PPCME PPI p-preserving PRES PRF PRO PROG PROH PROX PRS PRT PST PTCP PV Q QR QUD R RC REAL really-PosQ REC REFL REP RheQ RSL RT S SA SBJV

Phonological Form Polarity Item plural pleonastically negative parenthetical polarity, polarity head, polarity particle Polar-Interrogative Polarity Phrase positive clause positive question possessive (Chapter ) possibility (Chapter ) past perfect (Chapter ) Prepositional Phrase (Chapters , ) Penn-Helsinki Parsed Corpus of Early Modern English Penn-Helsinki Parsed Corpus of Middle English (nd edition) Positive Polarity Item presupposition preserving negation present perfect prominence (clitic) progressive prohibitive proximate present particle past participle preverb question quantifier raising Question Under Discussion relation relative clause realis Positive Question with really or VERUM focal accent recent reflexive repetitive Rhetorical Question Russian Sign Language Reaction Time Sentence (Chapter ) Speech Act subjunctive

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 SG SKCTC SMG SN SNE S-negation SOV SpecFocP SpecIP SpecNegP SpecTopP SpecVP SpecXP SSD SST STAT SUB SUBS SUBJ SubjP SVO T TAM TAM TDH θ-position TİD TMS TOP TopP TP u UE [uNEG] UP V V-movement VGT VP vP VP-internal whwh-exclamatives WYSIWYG

singular Southeast Kentucky Community and Technical College Supramarginal gyrus single negation Southern New England Sentence (sentential) negation Subject-Object-Verb Specifier of Focus Phrase Specifier of Inflection Phrase Specifier of Negation Phrase Specifier of Topic Phrase Specifier of Verb Phrase Specifier of XP, for some X Stop-Signal delay Stop-Signal task stative subordinate, subordinator substitution subjunctive Subject Phrase Subject-Verb-Object Tense tense-aspect-mood marker (Chapter ) Tense-Aspect-Mood verbal categories (Chapter ) Trace Deletion Hypothesis thematic position/theta-position Turkish Sign Language transcranial magnetic stimulation Topic Topic Phrase Tense Phrase utterance upward entailment, upward entailing uninterpretable negative formal feature Utterance Phrase verb Verb Second-movement Flemish Sign Language Verb Phrase phrase headed by the functional category v Verb Phrase-internal interrogative, relative exclamative sentences introduced by a wh-word what you see is what you get

xvii

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

xviii X0 XP ! 9 ∧ 8 © þ> ¬ Σ ΣP σ

 syntactic head syntactic phrase entails existential quantifier logical conjunction universal quantifier contrary negation implicates contradictory negation functional head encoding polarity features sigma phrase scalar

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

T  C ...................................................................

Katrin Amunts’s research tries to better understand the organizational principles of the human brain, by developing a multi-level and multi-scale brain atlas, and through the use of methods of high-performance computing to generate ultra-high resolution human brain models. She held a postdoctoral fellowship at the C. & O. Vogt Institute of Brain Research at Düsseldorf University, Germany, and then set up a new research unit for Brain Mapping at the Research Center Juelich, Germany. Currently, she is Professor of Brain Research, director of the C. and O. Vogt Institute of Brain Research, Heinrich-Heine University Düsseldorf and director of the INM-, Research Center Jülich. She is also a member and an elected vice chair of the German Ethics Council, the programme speaker of the programme Decoding the Human Brain of the Helmholtz Association, Germany, the Scientific Research Director and Chair of the Science and Infrastructure Board (SIB) of the European Union’s Human Brain Project, and a co-speaker of the graduate school Max-Planck School of Cognition. Since  she has been a member of the International Advisory Council Healthy Brains for Healthy Lives, Canada Jennifer Austin is a Professor in the Department of Spanish and Portuguese Studies at Rutgers University, Newark. She received her Ph.D. in Linguistics from Cornell University with a minor in Cognitive Science. Her research interests include language acquisition, bilingualism, and the effects of language contact, and she has published articles on the acquisition of Basque, English, and Spanish. She is also a co-author of the book Bilingualism in the Spanish-speaking World (Cambridge University Press, ) together with María Blume and Liliana Sánchez. David Beaver (Ph.D., University of Edinburgh, ) is a Professor in Linguistics and Philosophy at the University of Texas at Austin, and director of UT’s Cognitive Science Program. In addition to over sixty articles and book chapters on a wide range of topics in semantics, pragmatics, and beyond, he is author of the books Presupposition and Assertion in Dynamic Semantics (CLSI Publications, ), Sense and Sensitivity: How Focus Determines Meaning (Blackwell, , with Brady Z. Clark), and Hustle: The Politics of Language (forthcoming, Princeton University Press, with Jason Stanley). He is joint founding editor of the journal Semantics and Pragmatics. Frances Blanchette is Assistant Director of the Center for Language Science and Assistant Research Professor of Psychology at Penn State. Her  thesis investigated the syntax of English Negative Concord, Negative Polarity, and Double Negation, using theoretical and experimental methods. She has published papers on topics pertaining to negation in journals including Linguistic Variation, Glossa, Canadian Journal of Linguistics, and Language and Cognition. She also contributed to creating The Audio-Aligned and Parsed

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

xx

 

Corpus of Appalachian English (Tortora, Santorini, Blanchette, and Diertani ), and her current work includes a project to document and understand rural varieties of English in Central Pennsylvania. Manuel Bohn is currently a Marie Skłodowska-Curie Fellow at Stanford University and Leipzig University. His empirical and theoretical work is concerned with the psychological basis of human and great ape communication. Publications include Communication about Absent Entities in Great Apes and Human Infants (, Cognition) and Common Ground and Development (, Child Development Perspectives). His ongoing work focuses on the integration of social information during language learning in young children. Anne Breitbarth is a Senior Lecturer in Historical German Linguistics at Ghent University. Her research focuses on the determinants of syntactic change and diachronic stability, besides corpus linguistics, and West Germanic dialectology. She is the principle investigator of two infrastructure projects building parsed corpora (one of Middle Low German and one of southern Dutch dialects), and of an international research project on researching the influence of the syntax-prosody interaction on ongoing syntactic change. She has previously been a research fellow with the Flemish Science Foundation (FWO), at the University of Cambridge, and a research assistant at Tilburg University, where she completed her Ph.D. on auxiliary ellipsis in Early Modern German in . Josep Call is Professor in the Evolutionary Origins of Mind at the University of St Andrews (UK) and Director of the Budongo Research Unit at Edinburgh Zoo. His research focuses on elucidating the cognitive processes underlying technical and social problem solving in animals with the ultimate goal of reconstructing the evolution of human and nonhuman cognition. Recent publications include: Chimpanzees Consider Humans’ Psychological States when Drawing Statistical Inferences (, Current Biology) and APA Handbook of Comparative Psychology (, APA). His ongoing work focuses on the relation between coordination, communication, and social transmission. Ken Ramshøj Christensen is an associate Professor in Neurolinguistics at the English Department, School of Communication and Culture, Aarhus University. He is author or co-author of a wide range of peer-reviewed papers on syntax, negation, parsing, linguistic illusions, and language in the brain. Current projects include cross-linguistic variation and the processing of syntactic islands. Karen De Clercq is a postdoctoral researcher funded by the FWO and working at Ghent University. She wrote her Ph.D. on the nanosyntax of negative markers under the supervision of Prof. Liliane Haegeman. She is currently working on the fine-grained morphosyntax of Quantity-words (many/much; few/little), adjectives, degree comparison, negation, and V. Denis Delfitto is Full Professor of General Linguistics at University of Verona. From  to  he was Associate Professor at Utrecht University and a member of the Utrecht Institute of Linguistics. He published many contributions in international scientific journals and peer-reviewed volumes on the comparative syntax and semantics of Romance and

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 

xxi

Germanic languages, on the syntax and semantics of pronominal dependencies, on the syntax of reference in natural language, and on issues of language change and language impairment. His present research interests include issues at the interface between linguistics, the philosophy of language, and cognitive neuroscience. Kristin Denlinger is a Ph.D. student in Linguistics at the University of Texas at Austin. Her concentration is in syntax, semantics, and the interface between them, and she is currently working on a project related to the argument structures of denominal verbs. During her time as an undergraduate at Emory University, she worked in several language acquisition and cognition labs, and is currently pursuing a secondary focus in cognition. She is also interested in the linguistics of humor. Viviane Déprez is a native of Paris, France, who grew up in German-speaking Switzerland, and went to the United States to complete her Ph.D. in the Department of Linguistics and Philosophy at MIT. After graduating, she joined the Rutgers Department of Linguistics and also became a Research affiliate of the Cognitive Science Lab at Princeton University until , and subsequently an affiliate member of the Rutgers University Center for Cognitive Sciences and of the Rutgers Graduate Faculty of Psychology. She is currently a member of the CNRS Lab on Language, Brain, and Cognition in Bron and the director of the Comparative Experimental Linguistics (CELL) Lab at Rutgers. Isabelle Deschamps is a college Professor at Georgian College in Orillia, Ontario, Canada and a researcher in the Speech and Hearing Neuroscience Laboratory in Quebec City, Quebec, Canada. Her research focuses on the interaction between language processes (e.g. phonological) and other cognitive functions (e.g. memory and attention) in healthy adults and ageing. Henriëtte de Swart is Professor of French Linguistics & Semantics at Utrecht University. She has published widely on the cross-linguistic semantics of tense & aspect and indefinites, as well as negation, polarity, and negative concord. The monograph Expression and Interpretation of Negation: An OT Typology appeared in . In her project “Time in Translation” she is currently exploring parallel corpora as a new methodology to investigate cross-linguistic semantics. She is an associate editor of the journal Natural Language and Linguistic Theory. Carolin Dudschig received her Ph.D. in  after working for three years in the School of Psychology and Neuroscience at the University of St. Andrews. Her research has always been driven by the question of how the human brain deals with conflicts during information processing. Moving on from basic conflicts triggered during action control, her recent research investigates how the brain deals with conflicts originating during language comprehension, for example when processing negation. She was recently awarded a Heisenberg fellowship from the German Research Foundation to investigate the interrelations between linguistic and non-linguistic conflict processing. Veena D. Dwivedi is an associate Professor in Psychology/Neuroscience, and Director of the Dwivedi Brain and Language Lab, at Brock University, Canada. After her B.Sc. in

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

xxii

 

Physiology/Immunology from McGill University, she completed a Ph.D. in Linguistics from the University of Massachusetts-Amherst () examining syntactic and semantic dependencies in Hindi relative phrases. She has pioneered the development of experimental studies using formal semantic theory, starting with behavioral work in , then with Event Related Potentials (ERPs) starting in . Her current research focus is on the cognitive neuroscience of sentence processing, using ERPs. M.Teresa Espinal is Professor of Linguistics at the Universitat Autònoma de Barcelona, where she is a member of the Center for Theoretical Linguistics. Her main research interests are the theory of language, the syntax–semantics interface, and their relationship with a general theory of cognition. Her most recent research focuses on the structure and meaning of negation in natural languages, reference to kinds and to other generic expressions in Romance languages, the structure and meaning of bare nominals in Romance, and idiomatic constructions. Previous research dealt with different adverbial expressions and the adjunct/disjunct asymmetry. Anamaria Fălăuș is a CNRS researcher at the Laboratoire de Linguistique de Nantes, France. She received her Ph.D. from the University of Nantes in  and did postdoctoral research at the University of the Basque Country (UPV/EHU). Her work focuses on the semantics of natural language and its interface with pragmatics and syntax, from a crosslinguistic perspective. Her published work in these areas include an edited volume on alternatives in semantics (Palgrave Macmillan, ), as well as articles on polarity and free choice phenomena, epistemic indefinites, negative concord, epistemic future, and free relatives. Nicholas Fleisher is associate Professor of Linguistics at the University of WisconsinMilwaukee. His research to date has focused principally on the syntax and semantics of comparatives and degree constructions, scope and ellipsis in comparatives, and the English tough construction. His work has appeared in a variety of journals, including Language, Linguistic Inquiry, Semantics and Pragmatics, Journal of Semantics, and the proceedings of the North East Linguistic Society (NELS) and Semantics and Linguistic Theory (SALT). Naomi Francis is a Ph.D. candidate in the Department of Linguistics and Philosophy at the Massachusetts Institute of Technology. Martín Fuchs is a Ph.D. candidate in Linguistics at Yale University. He has worked in the morpho-syntactic properties of agrammatism. His current research examines the cognitive mechanisms of language processing that give rise to meaning variation and change. Anastasia Giannakidou is Professor of Linguistics at the University of Chicago, and codirector of the Center for Gesture, Sign, and Language. She is the author of many articles on the topics of negation, negative polarity, negative concord, free choice phenomena, quantification, definiteness, and modality—and author and editor of books such as Polarity Sensitivity as Nonveridical Dependency (John Benjamins), Definiteness and Quantification (Oxford), NP Structure in Slavic and Beyond (Mouton de Gruyter), and Mood, Tense, Modality Revisited (University of Chicago Press).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 

xxiii

Chiara Gianollo is Senior Assistant Professor of Linguistics at the University of Bologna. She obtained her MA and Ph.D. from the University of Pisa and has held appointments as lecturer and researcher at the Universities of Trieste, Konstanz, Stuttgart, and Cologne. Her main research areas are diachronic syntax and semantics, with specific focus on the use of formal theoretical linguistics to investigate the history of Greek, Latin, and Old Romance. She is the co-editor, with Agnes Jäger and Doris Penka, of Language Change at the Syntax– Semantics Interface (de Gruyter, ) and the author of Indefinites between Latin and Romance (Oxford University Press, ). Yosef Grodzinsky is Professor and director of the neurolinguistics laboratory at the Edmond and Lily Safra Center for Brain Sciences, the Hebrew University of Jerusalem, where he arrived in , after long service as a Senior Canada Research Chair of Neurolinguistics at McGill university. He is also a Visiting Professor at the Research Center Jülich. His research focuses on syntax and semantics and their neuroanatomical bases, topics which he has been studying through behavioral and neuroimaging experimentation, conducted on populations of healthy individuals, as well as on people with brain disease of various types. Jean-Rémy Hochmann is a tenured researcher at CNRS—Institut des Sciences Cognitives Marc Jeannerod in France. He has been investigating infant cognition since . Using mainly eye-tracking and EEG paradigms, he focuses on the developmental origins of human higher cognitive abilities such as language and reasoning. Laurence R. Horn is Professor Emeritus of Linguistics and Philosophy at Yale University. He is the author of A Natural History of Negation (Chicago, ; CSLI, ) and over  papers and handbook entries on negation, polarity, implicature, presupposition, pragmatic theory, word meaning, grammatical variation, and lying. His Ph.D. dissertation (UCLA, ) introduced scalar implicature. His six (co-)edited volumes include Negation and Polarity (OUP, ) and The Expression of Negation (de Gruyter, ). He is an elected fellow of the Linguistic Society of America and was past editor of the Outstanding Dissertations in Linguistics series (Garland/Routledge). Sabine Iatridou is Professor of Linguistics in the Department of Linguistics and Philosophy at the Massachusetts Institute of Technology. Virginia Jaichenco has a Ph.D. in Linguistics, and is Professor of Neurolinguistics and Psycholinguistics at the University of Buenos Aires. She works at the Institute of Linguistics of the same university where she leads a research team in these fields. She also teaches postgraduate courses of specialization in clinical neuropsychology. Her research deals with morphological and syntactic processing in children, adults, and brain injured patients. Shrikant Joshi obtained his doctorate in Linguistics from Université de Lausanne with doctoral research focusing on the semantics of affixation and its formalization. Prior to that, he obtained a Masters in French and a Bachelors in Electronics Engineering, University of Pune. His extensive experience is a blend of academia and industrial R&D in AI Systems. He has been teaching courses in Linguistics at the Department of Foreign Languages and courses in NLP at the Department of Technology at the University of

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

xxiv

 

Pune. His research interests are Morpho-semantics, Natural Language Understanding, and Computational Linguistics. Barbara Kaup’s research addresses the processes and representations involved in language comprehension, in particular with regard to semantic and pragmatic aspects of sentence interpretation. The focus of her research has always been the question of how negation is processed and represented during language comprehension. Barbara Kaup obtained her Ph.D. in Psychology from the Technical University of Berlin with a thesis that was prepared at the Graduate School of Cognitive Science at Hamburg University. Afterwards she spent three years as a postdoc in the lab of Don Foss at Florida State University, and then returned to Berlin to head an Emmy-Noether Group concerned with the processing and representation of negation during language comprehension. She is currently Professor for Cognitive Psychology at the University of Tübingen. Olga Krasnoukhova is a postdoctoral researcher at the University of Antwerp, Belgium. Her research focuses on different types of negation in South American indigenous languages by exploring the issue from typological, diachronic, and language contact perspectives. Olga received her Ph.D. from Radboud University Nijmegen with a thesis on Noun Phrase structure in the languages of South America. Her main research interests and publications cover topics such as negation from synchronic and diachronic perspectives, areal typology, typology of noun phrase, attributive possession, demonstratives, classifiers, and nominal number. Pierre Larrivée works on the semantics and pragmatics of grammar. He has published in journals such as Glossa, Lingua, Linguistics, Journal of Pragmatics, English Language and Linguistics, and Journal of French Language Studies on topics of negation, questions, indefinites, and expressives, in the synchrony and diachrony of French and English. His empirical focus on data reflecting the immediate competence of speakers has led him to initiate collective projects that have resulted in the creation and dissemination of diachronic and synchronic vernacular data corpora. He worked at Aston University (Birmingham, UK), and now holds a chair at Université de Caen (France). Yonatan Loewenstein received a B.Sc. in physics and a Ph.D. in computational neuroscience from the Hebrew University of Jerusalem in  and , respectively. Between  and , he was a Postdoctoral Fellow at MIT. Since , he has held a faculty position in the Departments of Neurobiology and Cognitive Sciences, the Edmond and Lily Safra Center for Brain Sciences, and The Federmann Center for the Study of Rationality, all at the Hebrew University of Jerusalem. Ana Maria Martins is Professor at the School of Arts and Humanities, University of Lisbon. Her research covers topics such as word order, clitics, negation, emphatic polarity, infinitival structures, and passive and impersonal constructions. She has directed projects resulting in parsed corpora for the study of the syntax of European Portuguese dialects (CORDIAL-SIN) and Old Portuguese. She has published in journals such as Lingua, Natural Language and Linguistic Theory, Linguistic Inquiry, and edited volumes by Oxford

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 

xxv

University Press, Benjamins, and de Gruyter. She is editor of Word Order Change (Oxford University Press, ), Manual de Linguística Portuguesa (de Gruyter, ), and co-editor of the journal Estudos de Lingüística Galega. Clemens Mayr received his Ph.D. from Harvard University in . He currently holds a position in formal semantics/pragmatics at the University of Göttingen. He has published, among other topics, on focus, implicatures, questions, intervention effects, and presuppositions. His main research question is how interpretive considerations constrain grammar. Jacques Moeschler is Professor at the Department of Linguistics, Geneva. His research has focused on tenses, causality, discourse, and logical connectives in French, as well as negation. Recent textbooks with Sandrine Zufferey include Initiation à l’étude du sens () and Initiation à la linguistique française (). His recent research articles have addressed formal and natural language, the pragmatics of logical connectives, the pragmatics of quantifiers, and the pragmatics of descriptive vs. metalinguistic negation. A textbook Implicatures (with Sandrine Zufferey and Anne Reboul) was published in  by Cambridge University Press. A monograph on Non-Lexical Pragmatics was also published in , by de Gruyter. Hanna Muller is a Ph.D. student in the Department of Linguistics at the University of Maryland. Her research interests are in psycholinguistics, semantics, and cognitive neuroscience. She holds a BA in Language and Mind from New York University. Liuba Papeo is a cognitive neuroscientist, tenured researcher at the Centre Nationale de la Recherche Scientifique in France and University Claude Bernard Lyon , and European Research Council laureate. Since , she has led the team Neurodev (Neuropsychology and Development) at the CNRS Institute of Cognitive Sciences Marc Jeannerod. In , she was awarded a European Research Council Starting Grant. Her research agenda, employing behavioral methods, neuroimaging, neurostimulation, and electrophysiology in human infants and adults, includes three main projects. The first addresses how the human brain realizes the amazing ability to readily detect and recognize conspecifics (human bodies and faces) in the cluttered environment. The second studies how humans understand others’ actions seen in isolation or in a social context. The third addresses the representations of communicative gestures and words, and how the two are organized within the brain network that represents meaningful signs. Elizabeth Pearce gained her Ph.D. at the University of Illinois in . She is an Honorary Senior Fellow at the University of Melbourne and was formerly a Senior Lecturer in Linguistics at Victoria University of Wellington. Her primary research interests are in syntax and historical linguistics. She has worked on aspects of the syntax of Romance languages and of Oceanic languages and she has carried out fieldwork in Vanuatu. She is the author of two books: Parameters in Old French Syntax: Infinitival Complements () and A Grammar of Unua (). Colin Phillips is Professor of Linguistics at the University of Maryland, where he also directs the Maryland Language Science Center. His research interests are in psycholinguistics, cognitive neuroscience, language acquisition, comparative linguistics, and in

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

xxvi

 

interdisciplinary graduate education. He holds a BA in Modern Languages from Oxford University and a Ph.D. in Linguistics from the Massachusetts Institute of Technology. Peter Pieperhoff studied physics in Cologne from  to . After having worked in industry in the field of automation technology, he began to investigate neuroscience in  in the Research Centre Jülich. His fields of interest are methodological developments for the analysis of imaging data, and to investigate the effects of ageing, neurodegeneration, and other types of diseases on the human brain. Cecilia Poletto is full Professor in the Department of Linguistic and Literary Studies at the University of Padua (part-time) and full Professor of Italian and French linguistics at the institute for Romance studies at the University of Frankfurt am Main (part-time). She is also an associate member at the Institute of Cognitive Sciences and Technologies—ISTC—(Padua). Her research interests include formal morpho-syntax of Romance varieties, in particular the Italian and French ones, and including non-standard varieties. She is currently a member of the steering committee for the graduate school in linguistic, phonological, and literary studies of the University of Padua, in charge of the Master in Linguistics, and a member of the DFG steering committee “Nominal Modification” at the Goethe University Frankfurt. Pilar Prieto is an ICREA Research Professor at the Department of Translation and Language Sciences at Universitat Pompeu Fabra, Barcelona, Catalunya. After obtaining a Ph.D. in Romance Linguistics at the University of Illinois at Urbana-Champaign (Illinois, USA), she was postdoctoral fellow at Bell Laboratories (NJ, USA). Her main research interests focus on the description of how prosody and gesture work in language and how they interact with each other and with other language components (pragmatics, syntax). She is also interested in how babies acquire sounds, melody, and gesture together with grammar, and how these components integrate in language development. Josep Quer is ICREA Research Professor affiliated to the Department of Translation and Language Sciences of the Pompeu Fabra University since , where he is the head of the Catalan Sign Language Lab (LSCLab). He has been Professor of Romance Linguistics at the University of Amsterdam, and an ICREA research Professor at the Department of Linguistics of the University of Barcelona. With the GRIN research group he developed the first grammatical description of LSC (). He gained a Ph.D. in linguistics from Utrecht University () and master’s degree in Grammatical Theory from the Autonomous University of Barcelona (). He has been the co-editor of the Sign Language and Linguistics journal since . David Ripley is a lecturer in philosophy at Monash University. His work centers on logic, language, and the relations between them. In particular, this includes a focus on nonclassical and substructural logics, and their potential applications to vague language and semantic paradoxes. Recent publications include “Experimental philosophical logic” in A Companion to Experimental Philosophy and “Blurring’ ” in the Notre Dame Journal of Formal Logic. Maribel Romero is Full Professor at Department of Linguistics of the University of Konstanz. Her main research interests lie in formal semantics and its interface with syntax

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 

xxvii

and pragmatics. She has worked on questions, focus, ellipsis, scope and reconstruction, specificational copular sentences, and degree constructions, with her current work focusing on epistemic bias, conditional sentences, discourse particles, and attitude verbs. Her list of publications includes contributions to Natural Language Semantics, Linguistics and Philosophy, Linguistic Inquiry, Natural Language and Linguistic Theory, Journal of Semantics and Glossa, among others. Liliana Sánchez is a Professor at Rutgers University. She has published Bilingualism in the Spanish-Speaking World with Jennifer Austin and Maria Blume (Cambridge University Press, ), The Morphology and Syntax of Topic and Focus: Minimalist Inquiries in the Quechua Periphery (John Benjamins, ), and Quechua-Spanish Bilingualism. Interference and Convergence in Functional Categories (John Benjamins, ) as well as articles in journals such as Bilingualism: Language and Cognition, Glossa International Journal of Bilingualism, Linguistic Approaches to Bilingualism, Lingua, Probus, and Studies in Second Language Acquisition. She is a co-founder of the Ph.D. Program in Bilingualism and Second Language Acquisition at Rutgers. María Elina Sánchez obtained her Ph.D. in Linguistics at the University of Buenos Aires, where she also teaches Neurolinguistics. She works at the Institute of Linguistics of the same university with a postdoctoral fellowship awarded by the National Research Council (CONICET). Her research focuses on morpho-syntactic processing in normal and brain damaged population. Barry Schein is Professor of Linguistics at the University of Southern California. He is the author of Plurals and Events (MIT Press, ) and ‘And’: Conjunction Reduction Redux (MIT Press, ) and contributor to The Oxford Handbook of Philosophy of Language () and The Routledge Companion to Philosophy of Language (). He has published articles in event semantics on event identity, plurals, reciprocals, definite description, conjunction, and negation. Rosalind Thornton is a Professor in the Linguistics Department at Macquarie University, Sydney, where she teaches child language acquisition. Rosalind has worked on a variety of topics including wh-movement, ellipsis, binding, quantification, and negation. She has worked on topics in negation across languages and in both typically developing children and children with specific language impairment. A current project investigates children’s knowledge of scope freezing in argument structure alternations in typically developing children. Christina Tortora is Professor of Linguistics at the City University of New York. She is a two-time National Endowment for the Humanities Fellow, and recipient of numerous grants from the National Science Foundation, to support the study of Italian dialects and grammatical variation in American English. She is the author of A Comparative Grammar of Borgomanerese () and Understanding Sentence Structure: An Introduction to English Syntax (), and has published numerous papers in edited collections and in journals such as Linguistic Inquiry, Natural Language and Linguistic Theory, Theoretical Linguistics,

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

xxviii

 

and American Speech, on syntactic theory, micro-syntactic variation, and intra-speaker variability. Lucia M. Tovena is Professor of Linguistics at the University Paris VII, specializing in formal semantics. Her areas of interest include negation and forms of individuation in the nominal and verbal domain. She has worked on epistemic determiners, NPIs and freechoice items, negative concord, neg-raising, aspect and pluractionality associated with evaluative morphology, event nouns and the notion of nomen vicis. She is the author of The Fine Structure of Polarity Sensitivity (). She led the project ELICO Corpus of annotated occurrences of French determiners (), and edited the volume French Determiners in and Across Time (). She is currently working on distributivity and ratio expressions. Susagna Tubau is Associate Professor at the Universitat Autònoma de Barcelona, where she teaches English syntax, as well as other general linguistics courses. Her research revolves around the expression of negation in English, Catalan, Spanish, and, most recently, Basque. Her major publications are on Negative Concord in Romance and in non-standard varieties of English, Double Negation in Catalan and Spanish, English minimizers, the nonstandard English pragmatic particle innit, and yes- and no-answers to negative yes/noquestions. Johan van der Auwera is Emeritus Professor of General and English Linguistics at the University of Antwerp. He was elected as the President of the Societas Linguistica Europaea in  and as member of the Academia Europaea in . His research focuses on grammatical semantics with special reference to conditionals, mood, modality, negation, indefinites, impersonals, and similatives, from a synchronic and diachronic as well as an areal perspective, and occasionally from a historiographical one. Languages studied are English, including New Englishes and Creoles, Germanic languages, European languages, and samples or interesting selections of the world’s languages. Manuel de Vega is Professor of Psychology and Neuroscience and Director of the Instituto Universitario de Neurociencia, a multidisciplinary research center, at the Universidad de La Laguna, Spain. His research combines behavioral neuroimaging, electrophysiological methods, and non-invasive brain stimulation techniques, in order to address the general issue of how motor, emotional, and visuo-spatial neural networks could be recycled to process linguistic meaning. He is also involved in the study of the neural bases of discourse comprehension, especially of narrative texts. Christoph J. Völter is currently a postdoc at the University of Veterinary Medicine, Vienna. His empirical and theoretical work is concerned with the evolutionary origins of flexible behaviour and abstract thought. Publications include Comparative Psychometrics: Establishing what Differs is Central to Understanding what Evolves (, Philosophical Transactions of the Royal Society B) and Great Apes and Children Infer Causal Relations from Patterns of Variation and Covariation (, Cognition). His ongoing work focuses on

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 

xxix

the structure of individual differences in domain-general cognitive abilities such as inhibitory control and working memory. Phillip Wallage is Senior Lecturer in English Language and Linguistics at Northumbria University. He specializes in morpho-syntactic variation and change, in particular the application of quantitative and statistical methods to model changes in progress within diachronic corpus data. He has published widely on the syntax and pragmatics of early English negation, in journals such as Lingua and English Language and Linguistics. His most recent publication is a monograph on early English negation for Cambridge University Press. His current work investigates the diachrony of polar responses in English and other languages. Andrew Weir is Associate Professor of English linguistics at the Norwegian University of Science and Technology, Trondheim. He received his Ph.D., on the syntax and semantics of fragment answers and clausal ellipsis, from the University of Massachusetts Amherst in . In addition to ellipsis and issues of the phonology–syntax–semantics interfaces, he is interested in register variation (e.g. the grammatical properties of diaries, recipes, and other ‘reduced’ written registers), on which topic he has published articles in Linguistic Variation and English Language and Linguistics, and (micro)comparative syntax within Germanic, especially dialects of English. Hedde Zeijlstra is currently Professor of English Linguistics and theory of grammar at the University of Göttingen, Germany. Previously, he held positions at the universities of Amsterdam and Tübingen, and at MIT. He has worked intensively on negation and negative dependencies, including negative concord, negative polarity sensitivity, and positive polarity sensitivity. He is the author of Negation and Negative Dependencies, which will appear with Oxford University Press.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

  ......................................................................................................................

 Negation in language and beyond ......................................................................................................................

    . 

.. M

.................................................................................................................................. F centuries, negation has been a topic that has fascinated logicians, philosophers, linguists, and psychologists alike. It relates to a whole set of notions, that include opposition, falsity, absence, non-existence, denial, rejection, refusal, correction, avoidance, disappearance, prohibition, and the like. Some of these notions have been at the center of important debates within various disciplines, while others have only served to feed intuitions in the study of language and mind. Negation offers a privileged and still wide-open window into the workings of the human mind. In its simplest form of mere refusal, already rooted in non-verbal communication, negation surely is not unique to mankind, but as soon as one takes the further step of contemplating difference, absence, or non-existence, as concepts, and not the mere observation of a disappearance, it quickly becomes questionable whether or not humans are the only species capable of such a mental feat. The question rapidly turns to certitude when the contemplation moves to denial. When truth and falsity become the heart of the matter, recursive discreetness is reached, and with it, formal capacities that are yet to be discovered in other species. As Larry Horn (b: ) puts it, “negation is a sine qua non of every human language but is absent from otherwise complex systems of animal communication” and, hence, in part “is what makes us human, imbuing us with the capacity to deny, to contradict, to misrepresent, to lie and to convey irony.” What cognitive capacities are involved in each of these steps is a question that a number of neuroscientists and psychologists are beginning to investigate, and that has become increasingly pertinent to linguists and psycholinguists asking how language evolved, if it did, and how negation is acquired by children from the preverbal to the verbal stage. Research in the development, processing, and understanding of negation by various populations, and at various stages of cognitive faculties is becoming increasingly relevant to linguists and philosophers while, at the same time, experimental and psycholinguistic research cannot flourish without the nourishing of theoretical linguistics and its results. Furthermore, the growing availability of global exchange and of new ways of electronically acquiring and storing collected data, fueled by the current urgency of recording

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



   . 

language diversity before it diminishes, enables access to comparative linguistic data on so far poorly explored terrain that can either solidify or question current empirical generalizations. Digitalization of existing historical data has also opened new possibilities of exploring and synthesizing, confirming or revisiting historical evolution, for which the well-known Jespersen negation cycle has provided a pillar that has anchored many theoretical models of diachronic change. With this in mind, research on negation is as rich and alive as ever, and there still exist fundamental questions that remain to be discovered. Among these central questions, we would like to mention the following. First, if negation is so fundamental in the cognition of human beings, why does it have so many forms? Besides the many forms that negation can take cross-linguistically, it is also the case that human beings can express negation in nonlinguistic bodily-based gestures sometimes used to accompany the linguistic expressions and sometimes used to replace them. Second, what are the factors that intervene in the interpretation of a negative form? It appears to be the case that the nature of expressions that can express negation is diverse among natural languages, which provides evidence that the grammar of natural languages show variation in the expression of negation. Third, how do prosody and gesture interact in the interpretation of negative utterances? Recent studies have highlighted the precursor role of gesture in the early development of negation and the need of an integrated analysis within a multimodal interaction model of language design. These are only some of the questions that will be approached in the Handbook of Negation. Besides the enduring landmark reference of Larry Horn’s () Natural History of Negation there have been a number of recent volumes on the topic of negation that offer a synthesis of current theoretical and empirical linguistics research on negation. In this respect we would like to mention (i) Larry Horn’s (b) The Expression of Negation, a landscape of the cross-linguistic diversity of the forms of the expression of negation in natural languages; (ii) Henriette de Swart’s () Expression and Interpretation of Negation, a broad study of the markedness of negation and a typology of negative items from an Optimality Theory perspective; (iii) Pierre Larrivée and Chungmin Lee’s () Negation and Polarity: Experimental Perspectives, a comprehensive volume that houses together some increasingly interacting trends of research on negation, presupposition, implicature, and negative polarity. The ambition of the present Handbook is to offer a complete stateof-the-art of the study of negation from multiple perspectives and disciplines by assembling representative chapters by world specialists on this topic.

.. O

.................................................................................................................................. Part I is devoted to the discussion of fundamental philosophical questions: what is negation, denial, opposition, etc., and what types of negation are there? Part II focuses on the syntax of negation and addresses the following linguistic questions: how negation works in the syntax of natural languages, and what are the main structural factors that have to be taken into account in the study of negation.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi





Part III deals with fundamental questions that involve negation at the interface between syntax and semantics: neg-raising, intervention effects with negation, the form and function of negative questions, expletive negation, and the scope of negation. Part IV addresses the main topics at the interface between the semantics and the pragmatics of negation: does negation always correspond to a propositional operator? What is the interaction of negation with other operators? Why do natural languages also require ways to express metalinguistic negation? How does negation relate to presupposition? Part V focuses on various types of negative dependencies in natural language. It aims to answer the question of what it means to be negative or to be sensitive to negation. Hence, it investigates the requirements that an item must fulfill to convey a negative meaning, the different types of negative dependent elements, and the form of these dependencies, the constraints they must obey and the factors that influence the interpretation of multiple dependencies with either a single negation or a double negation reading. Part VI focuses on synchronic and diachronic cross-linguistic variation in negation. It introduces corpus studies and data mining. It deals with the evolution of negation, the negative cycle, and the role of pragmatics in negation change. Part VII is devoted to the discussion of some fundamental psycholinguistic questions on the emergence and acquisition of negation: what the precursors of negation are in nonhumans, how and when negation emerges in preverbal infants and in children, as well as how negation is acquired in L. Finally, Part VIII introduces some novel perspectives and questions that emerge from neuroscience and neurobiology for the study of negation. The main question investigated is how linguistic negation can both be a window into the study of the brain and inform us more generally about the neurology of language.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ........................................................................................................................

FUNDAMENTALS ........................................................................................................................

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ......................................................................................................................

   Contradiction and contrariety in logic and language ......................................................................................................................

 . 

.. N     

.................................................................................................................................. I search of the essential properties of a theory of negation and opposition, we find that, as Zeno Vendler wrote in a different connection (b: ), “as it so often happens . . . we suddenly realize that the path of inquiry we hoped to open is already marked by the footprints of Aristotle.” Beginning in the Categories, Aristotle distinguishes four varieties of opposition, three of which inherently involve negation: contradiction, contrariety, privation, and correlation. Any two contradictory opposites, such as Grass is white/Grass is not white, divide the true and the false between them. Of two contradictories one is true if and only if the other is false: every substance is either white or not white but not both. Contrariety, on the other hand, involves opposition between terms that are mutually incompatible but not mutually exhaustive. Two contraries, for example Grass is white/ Grass is black, cannot be simultaneously true but they can be simultaneously false; they allow for an unexcluded middle, a substance neither white nor black. In privation, a privative term (X is unmusical; Y is blind) signals the lack of a property possessed by the corresponding positive (X is musical; Y is sighted). Privation, which can be seen as a special instance of contrariety, is typically realized in English by affixal negation, including the productive formation of un-nouns, as will be discussed below. Aristotle’s fourth kind of opposition is that between mutual converses, for example parent/child or own/belong to: X is a parent of Y if and only Y is a child of X; X owns Y if and only Y belongs to X. Since this opposition is not directly linked to negation, we will not discuss it further in this chapter. Other types of semantic opposition, such as that between clauses linked by adversative or concessive particles (but, although, while), will likewise be omitted from our discussion.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 . 

The focus of this chapter is the delineation of the first three kinds of semantic opposition and their reflexes in the morpho-syntax of natural language. Section . explores the nature of contradictory and contrary opposition as represented by the Square of Opposition. In section ., we delineate the correspondences between the contradictory negation and wide-scope (or sentence) negation as variously defined by Jespersen (), Klima (), and Jackendoff () and between contrary negation and the notion of constituent negation in grammatical theory. Section . provides a closer look at the prevalence of contrary opposition in natural language, and in particular at the tendency in a variety of contexts for formal contradictory negation to strengthen, semantically or pragmatically, to contrariety. The two primary axioms of Aristotle’s logic are the law of non-contradiction (LNC) and the law of excluded middle (LEM). In the Metaphysics LNC is defined as follows: “It is impossible that the same thing can at the same time both belong and not belong to the same object and in the same respect, and all other specifications that might be made, let them be added to meet local objections” (b–; cf. Horn  for complications). LNC lies at the heart of Aristotle’s theory of opposition, governing both contradictories and contraries. What distinguishes the two forms of opposition is a second indemonstrable principle, the law of excluded middle (LEM): “Of any one subject, one thing must be either asserted or denied” (Metaphysics b). These laws can be formalized, with some loss of faithfulness, in the standard propositional versions in (a) and (b) respectively, ignoring Aristotle’s stipulated modal and temporal modifications: ()

a. LNC: ¬(ϕ ∧ ¬ϕ) b. LEM: ϕ ∨ ¬ϕ

Both laws pertain to contradictories, as in a paired affirmation (‘S is P’) and denial (‘S isn’t P’): the negation is true whenever the affirmation is false, and the affirmation is true whenever the negation is false. Thus, a corresponding affirmation and negation cannot both be true, by LNC, but neither can they both be false, by LEM. But while LNC applies both to contradictory and contrary oppositions, LEM holds only for contradictories: “Nothing can exist between two contradictories, but something may exist between contraries” (Metaphysics b): a dog cannot be both black and white, but it may be neither, if it’s gray or brown. For Aristotle, contradiction, unlike contrariety, is restricted to statements or propositions; terms are never related as contradictories, both because only statements (subjectpredicate combinations) can be true or false (Categories b–) and because any two terms may simultaneously fail to apply to a given subject. This is obvious in the case of mediate contraries, which allow an unexcluded middle, as in the pairs in () and (). ()

a. This dog is white. b. This dog is black.

()

a. Socrates is happy. b. Socrates is sad.

But oppositions like that in ()

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   ()



a. Socrates is sick. b. Socrates is healthy.

are not contradictories either for Aristotle but immediate contraries, since here too both statements may both be false, even though every existing person is either sick or healthy: “For if Socrates exists, one will be true and the other false, but if he does not exist, both will be false; for neither Socrates is sick nor Socrates is healthy will be true, if Socrates does not exist at all” (b–). Category mistakes too yield immediate contraries: my pet rock Socrates is neither well nor ill. But given a corresponding affirmation and negation, one will always be true and the other false even if the subject referent fails to exist: “for if he does not exist, ‘he is sick’ is false but ‘he is not sick’ true” (b–). Another category of apparent contradictories that arguably constitute Aristotelian contraries is that of future contingents, statements about future events whose truth value cannot be assigned in the present, such as those in (a,b) ()

a. There will be a sea-battle tomorrow. b. There will not be a sea-battle tomorrow. c. Either there will be a sea-battle tomorrow or there won’t be.

While LNC is operative, preventing (a) and (b) from both being true, on one reading of Chapter  of Aristotle’s De Interpretatione LEM does not apply here, since (a) and (b) can both fail to be true today (cf. e.g. Lukasiewicz ; Anscombe ). Their disjunction in (c), however, remains necessarily true. One approach to this apparent paradox is by invoking the theory of supervaluation (see Sorensen : §; Horn : §). A linguistically more salient class of cases in which a sentence and its negation can both be judged false (or at least not true) is that of vague predications (Alxatib and Pelletier ; Ripley a). Thus, when evaluated in a context in which John is just short of  feet tall, (a) and (b) may both be judged false. ()

a. John is tall. b. John is not tall.

More generally, when F is a borderline instance of the application of predicates like bald, rich, or tall, both the affirmative predication a is F and its apparent contradictory negation a is not F may be rejected, although in the same circumstances speakers may also be willing to affirm that a is both F and not-F (John is tall and not-tall), which complicates matters. We have seen that in the Aristotelian system, terms like ‘sick’ and ‘healthy’—or ‘white’ and ‘not-white’—are immediate contraries, not contradictories, since neither term in each opposition can be affirmed of non-existent objects and neither applies to those items (e.g. abstract concepts) falling naturally outside the range of objects to which the predicate applies: Santa Claus and sincerity are neither tenured nor non-tenured. In this chapter, however, we will henceforth adopt the standard practice in linguistics of reclassifying such immediate contraries as contradictory opposites. In linguistic semantics (cf. Jespersen ; Lyons ) sentences like X is male/X is female; Y is white/Y is non-white and, derivatively, the relevant predicate terms (male/female; alive/dead; white/non-white) are taken to be

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 . 

contradictories rather than (as for Aristotle) immediate contraries, since the vacuous (non-denoting) subjects and category mistakes are disregarded for reasons of convenience. At least in the unmarked case, negation (S is not P) can be co-defined with affirmation (S is P) for a given logical subject/predicate pair S and P: An affirmation is a statement affirming something of something, a negation is a statement denying something of something . . . It is clear that for every affirmation there is an opposite negation, and for every negation there is an opposite affirmation . . . Let us call an affirmation and a negation which are opposite a contradiction. (Aristotle, De Interpretatione a–)

But this criterion, satisfied straightforwardly enough in the case of simple subject-predicate statements, must be recast for quantified expressions, whether universal (Every cat purrs; No cat purrs) or existential (Some cat purrs; Not every cat purrs). For such cases, Aristotle shifts from a formal to a semantically based criterion of opposition (De Int. b–, a–): [W]ith regard to particulars, if it is true when asked something, to deny it, it is true also to affirm something. For instance: ‘Is Socrates wise? No. Then Socrates is not-wise.’ With universals, on the other hand, the corresponding affirmation is not true, but the negation is true. For instance: ‘Is every man wise? No. Then every man is not-wise.’ This is false, but ‘then not every man is wise’ is true. This is the opposite [= contradictory] statement, the other [every man is not-wise] is the contrary.

This is best visualized in terms of the Square of Opposition developed by the later commentator Apuleius (for elaboration, see Horn : ch.; Parsons ), shown in Figure .. Every S is P A

subalterns

I Some S is P

contraries

contradictories

subcontraries

No S is P E

subalterns

O Some S is not P

 .. Traditional Square of Opposition

Members of an A/O pair as in () or an I/E pair as in () are contradictories, ()

a. Every woman is white. b. Not every woman is white.

()

a. Some woman is white. b. No woman is white.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



since in any state of affairs one member of each pair must be true and the other false. Members of an A/E pair like that in () ()

a. Every woman is white. b. No woman is white.

constitute contraries, since these statements cannot both be true simultaneously but can both be false. The contradictories of these contraries (Not every woman is white/Some woman is white) can be simultaneously true with reference to the same subject (b–). This last opposition of I and O statements, later to be dubbed subcontraries because they appear below the contraries given the geometry of the traditional square, is a peculiar opposition indeed. Aristotle (Prior Analytics b–) sees I and O statements as “only verbally opposed,” given the consistency of the I statement, for example “Some Greeks are bald,” with the corresponding O statement, “Not all Greeks are bald” (or “Some Greeks aren’t bald,” which doesn’t necessarily amount to the same thing, especially if there are no Greeks; see Parsons  on the problems of existential import). Thus, the distinction between contradictory and contrary opposition is ultimately a semantic one, based on the operation of LNC (applying to both contradictories and contraries) and LEM (applying only to contradictories). In an ideal world, the distinction between the two varieties of opposition would map directly onto the morpho-syntactic differentiation between sentential (or sentence) negation and subsentential (constituent) negation in natural languages. But as always the world is not ideal; the more complex mapping we actually find is addressed in section ..

.. C    

.................................................................................................................................. What is the range of possible mappings between contradictory opposition and sentential negation? If we think of negation as essentially a means for opposition—the impossibility of simultaneously endorsing two incompatible options—propositional negation is not necessarily privileged. This view is formally implemented in the Boolean algebraic model of Keenan and Faltz, on which negation is a cross-categorial operation (see Collins and Postal  for a recent implementation), as are the binary connectives: We can directly interpret conjunctions, disjunctions, and negations in most categories by taking them to be the appropriate meet, join, and complement functions of the interpretations of the expressions conjoined, disjoined, or negated. The sense in which we have only one and, or, and not is explicated on the grounds that they are always interpreted as the meet, join, and complement functions in whatever set we are looking at. (Keenan and Faltz : )

Treatments of English and other languages frequently posit negative operators whose scope is narrower than the sentence or clause. This tradition dates back to Aristotle, for whom the predicate term negation in Socrates is not-wise, affirming that the predicate not-wise holds

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 . 

of Socrates, yields a false statement if Socrates does not exist, while the predicate denial Socrates isn’t wise denies that the predicate wise holds of Socrates and is true if Socrates does not exist. For Jespersen (), the subclausal “special” negation as in Nobody came, where “the negative notion . . . belong[s] logically to one definite idea,” is opposed to “nexal” negation, applying to “the combination of two ideas,” typically the subjectpredicate nexus. Later linguists usually follow Klima () and Jackendoff () in allowing for constituent negation (e.g. verb phrase negation in You can [not go]) alongside sentential negation (You cannot go), utilizing various grammatical and semantic diagnostics for distinguishing the two varieties. Within the categorical term-based logic of Aristotle, every statement—whether singular or general (quantified)—is of subject-predicate form. Contradictory negation on this view is not a one-place connective taking propositions into propositions but rather a mode of predication, a way of combining subjects with predicates: a given predicate can be either affirmed or denied of a given subject. Predicate denial, while yielding the familiar truthconditional semantics of contradictory opposition, does not apply to its own output and hence does not syntactically iterate. In this respect, predicate denial prefigures Montague’s treatment of English as a formal language (Montague ; cf. Horn : §.), while also providing a more plausible representation of contradictory negation in natural language. Whether in Ancient Greek, in modern English, or elsewhere, reflexes of propositional negation, a sentence-external iterating one-place connective, are hard to find outside of artificial constructs like the ‘it is not the case’ construction. But even a single sentence-external negation mapping literally onto ¬ϕ as in the Stoics’ Not: It is day is a logician’s construct rarely attested in the wild (Geach ; Katz ; Horn ): The negation Aristotle was interested in was predicate negation; propositional negation was as foreign to ordinary Greek as to ordinary English, and he never attained to a distinct conception of it. The Stoics did reach such a conception, but in doing so they violated accepted Greek usage; their use of an initial oukhi a[s] the standard negation must have read just as oddly as sentences like “Not: the sun is shining” do in English. (Geach : )

Rather, contradictory negation is variably expressed as a particle associated with a copula or a verb, as an inflected auxiliary verb, as a verb of negation, as a negative suffix or prefix, but not as a one-place t/t operator.¹ As noted in section ., Socrates isn’t healthy is true in Aristotelian semantics when Socrates is ill, when Socrates doesn’t exist, or when Socrates is my pet rock and thus incapable of being healthy or ill (= not-healthy). But in addition to predicate denial, in which a predicate F is denied of a subject a, Aristotelian logic allows for narrowscope predicate term negation, in which a negative predicate not-F is affirmed of a. The relations of predicate denial and predicate term negation to a simple affirmative proposition and to each other can be schematized on a generalized square of opposition for

¹ For Horn (), apparent instances of external negation are taken to represent the metalinguistic use of negation. Bar-Asher Siegal (b) presents evidence for the existence of a semantic external negation operator in Jewish Babylonian Aramaic.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



singular (non-quantified) statements. The blueprint for such a square is provided by Aristotle (De Interpretatione b–, Prior Analytics I, ch. ), as shown in Figure .. S is P A

contraries

S is not-P E

entails

contradictories

entails

I

subcontraries

O

S isn’t not-P

S isn’t P

 .. Singular Square of Opposition

For immediate contraries formed by narrow-scope predicate term negation, for example white/not-white, the quasi-English a is not-F rendering corresponds to what is expressed in ancient Greek through marked word order, as in the distinction between for example einai mê leukon ‘to be not-white’ and mê einai leukon ‘not to be white’ (Prior Analytics I b). The difference between denying P of S and affirming not-P of S is realized as a scopal distinction: S P [not is] (Socrates healthy not-is) vs. S [not P] is (Socrates not-healthy is). For Aristotle, a is neither F nor not-F can be true if a doesn’t exist (Santa is neither white nor not-white) or isn’t the kind of thing that can be F (The number  is neither white nor notwhite), given that not-F is taken to affirm the negative property non-F-ness of the subject rather than denying a positive property. For any object x, either x is red or x isn’t red, but x may be neither red nor not-red—if, for instance, x is a unicorn or a prime number. This system of dual negations is not only internally consistent but empirically grounded; its echoes are recognizable in Jespersen’s () distinction between nexal negation (not happy) and special negation (unhappy), Von Wright’s () distinction between weak (contradictory) vs. strong (contrary) negation, and Jackendoff ’s () semantic reanalysis of Klima’s () grammatical categories of sentence vs. constituent negation. In each case, a negative marker whose scope is narrower than the proposition determines a statement logically distinct from a simple contradictory.² As these observations suggest, the linguistic tradition has typically invoked a syntactic rather than semantic definition of sentential negation. As Klima () points out, an important syntactic correlate of the distinction between wide-scope (sentential) vs. narrowscope (constituent) negation in English relates to word order, in particular the ability for fronted items to license subject-auxiliary inversion—often termed negative inversion, although as with negative polarity items (NPIs) it’s not just explicitly negative elements that serve as licensers (see Chapter  in this volume). As has long been recognized (Klima : ; Jackendoff : §.; Horn : ; Haegeman ; Büring , Collins and ² Related to this distinction is the often assumed—and often rejected—ambiguity between internal (choice) and external (exclusion) negation interacting scopally with putative semantic presuppositions; see, among many others, Atlas (), Horn (), and Carston (: ch. ) for extensive discussion and references.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 . 

Postal ), only when the “negative” element has sentential (wide) scope can it trigger inversion and license NPI ever, as in (a), (a), and (a). In the corresponding canonical word order versions, the scope of negation does not extend beyond the fronted phrase and NPI ever is excluded. ()

a. With no job will I be happy. b. With no job I will be happy.

()

a. In no clothes does Dana (ever) look good. b. In no clothes Dana (*ever) looks good.

()

a. At no time were we (ever) alone together. b. In no time we were (*ever) alone together.

Klima () proposed a set of diagnostics to distinguish sentential (wide-scope) and constituent (narrow-scope) negation, and other syntactic tests have since been added. The four Klima diagnostics are based on the correlation between tag formation and sentence polarity. As seen in () and (), the presence of sentential negation is marked by the possibility of either (vs. too) tags, neither (vs. so) tags, not even tags with pleonastic negation, and positive opposite polarity tag questions. Sentences containing constituent rather than sentential negation pattern with affirmative sentences in the distribution of these items. ()

a. b. c. d.

Priests cannot marry, and nuns cannot marry either. Priests cannot marry, and neither can nuns. Priests cannot marry, not even if they’re in love. Priests cannot marry, can they?

()

a. b. c. d.

Priests are unable to marry, and nuns are unable to marry {too/*either}. Priests are unable to marry, {and so/*neither} are nuns. Priests are unable to marry, {even/*not even} if they’re in love. Priests are unable to marry, {aren’t they/#are they}?

(While a positive tag is possible in (d), it represents the same-polarity tag used for challenging a speaker’s claim rather than seeking confirmation; cf. Cattell ). Another diagnostic for sentential negation is provided by the distribution of parentheticals. As first pointed out by Ross (), negation is permitted pleonastically within some parentheticals (members of the “neg-raising” class of opinion/belief; see Chapter  in this volume), so the two versions of (a) are synonymous, while the negative form of the parenthetical is ruled out by the constituent negation in (b) and (c): ()

a. It isn’t possible, I (don’t) think. b. It is impossible, I (*don’t) think, for priests to marry. c. Priests are unable, I (*don’t) think, to get married.

As seen in the contrast in (), such pleonastically negative parentheticals are only possible after negative force is marked by main clause sentential negation:

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   ()



a. It is not, I (don’t) think, possible for priests to marry. b. It is, I (*don’t) think, not possible for priests to marry.

Given the correlation between sentence negation and wide scope, we predict that negative/ modal structures will pass the tests for sentence negation if and only if the negation scopes outside the modal. In some cases, this prediction is clearly on target. Note that the negation in cannot and can’t can only take wide scope, while can not can be interpreted with narrow scope (= ‘possible not’) as well as wide scope (= ‘not possible’) negation (see Horn ,  for a proposed explanation). ()

a. b. c. d.

Priests cannot marry, and {neither/*so} can nuns. Priests can’t marry, {can/*can’t} they? Priests can (if they wish) [not eat meat], and {so/*neither} can nuns. Priests can (if they wish) [not eat meat], {can’t/*can} they?

However, while generally converging on intuitively correct predictions, these tests have been argued to be insufficient in deciding the crucial cases in that they often yield conflicting results; cf. Jackendoff (, ), Attal (), Ross (). Ross, for example, cites the data in (a,b) (his judgments): ()

a. ??He hardly damaged the car, did he? b. He hardly damaged the car, {?and neither did you/not even by filling it with gravy}. c. Nobody saw John, {did they/*didn’t they}? d. John saw nobody, {didn’t he/?did he}?

The variables determining negative strength cited by Ross include the degree of overtness of the negative element and the grammatical relation of the nominal containing the incorporated negation. As (c,d) show, quantifier negation in subject position functions as “more negative” than that in non-subject position, although both can be viewed as contradictory negations, of Somebody saw John and John saw somebody respectively. To illustrate the point on overtness of negation (Horn , Ross ), note that while few and not many are intersubstitutable salve veritate, they are not always intersubstitutable salva grammaticalitate. The essential difference between them is that not many is, while few is not, morphologically negative, a factor relevant for the behavior of these expressions with respect to the Klima diagnostics and negative parenthenticals. Similar observations apply to morphologically non-negative exclusive and quantificational adverbs, as opposed to their n-initial counterparts: ()

a. b. c. d.

{Not many/*Few} of my friends will be joining us, not even for dessert. {Not many/#Few} Democrats will support the bill, will they? {Not often/*Seldom/*Rarely} do you wear a seat belt, I don’t think. {Nobody but/*Only} Trump, I don’t suppose, would have said that.

Crucially, the relevant operator, whether overtly or covertly negative, provides a contradictory negation of the corresponding affirmative (Many of my friends will be joining us, You

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 . 

often wear a seat belt, etc.). This suggests a need to dissociate the grammatical property of sentential negation from the semantic notion of contradictory opposition. As we touched on above, Jespersen differentiates nexal and special negation. Nexal negation is clausebased, marked in the auxiliary in English, while special negation has clause-internal scope and is typically marked by a negative (lexically incorporated or not) which immediately precedes or is part of the element on which it focuses. While Klima’s distinction between sentence and constituent negation is not co-extensive with Jespersen’s, both are crucially syntactic in definition and spirit and thus do not necessarily reflect Aristotelian contradictory negation or the semantically based notions that follow Jackendoff ’s characterization of sentential negation: ()

A sentence [s X-neg-Y]s is an instance of sentence negation if there exists a paraphrase It is not so that [sX-Y]. (Jackendoff : )

To see how the criteria differ, note that in the classic minimal pair in (), ()

a. Not many of us wanted the war. b. Many of us didn’t want the war.

(a) is a case of special (constituent) negation for Jespersen (: ), while the auxiliary negation in (b) counts as nexal. But for Jackendoff, (a) is a more likely candidate for sentence negation than (b), since only the former can be paraphrased by It is not so that many of us wanted the war. Here as elsewhere, Klima’s diagnostics turn out to be indecisive in determining which (if either) of these two negatives is to be assigned to the class of sentence negations (cf. Attal ; Ross ). Jackendoff ’s semantic criterion for S-negation, essentially based on the truth-value-toggling feature of contradictory negation standard since Aristotle, is not only more straightforward to operationalize, but also more easily universalizable than the Klima-style English-based syntactic criteria utilizing position and distributional diagnostics. But it too leads to some odd results. As Attal (: ) points out, (a), like its French counterpart, ()

a. I don’t want to leave. [Je ne veux pas sortir.] b. I want to leave. [Je veux sortir.]

has a salient “neg-raised” interpretation (see Chapter  in this volume) on which it is not a contradictory but a contrary of its affirmative counterpart in (b): (a) is not equivalent to ‘It is not so that I want to leave’, but rather to ‘I want to not-leave’. Jackendoff ’s criteria would thus banish (a) from the ranks of S-negations; yet on both intuitive and syntactic grounds it ranks with undoubted sentential negations like I don’t have to leave, as Jespersen’s and Klima’s syntactically oriented criteria predict. We saw above in () that E-vertex modals with negation outscoping possibility (can’t, cannot) pattern with sentential negations (predicate denials) rather than constituent (term) negations; this is unproblematic, given their status as contradictory negations of the corresponding I-vertex possibility modals. Significantly, though, inflected negations of A-vertex (weak or strong) necessity modals like mustn’t or shouldn’t also pattern as

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



sentential negations, despite the fact that semantically they represent contraries rather than contradictories of their affirmative counterparts with should and must. ()

a. You shouldn’t buy that car, and {neither/*so} should your sister. b. You mustn’t come here again, and your brother mustn’t {either/*too}. c. He mustn’t have left yet, I don’t think.

If, as it appears, a negatively inflected auxiliary always constitutes a sentential negation, or in Aristotle’s terms a predicate denial, then predicate denials—like sentence negations à la Klima—do not invariably represent sentence negations à la Jackendoff, that is, sentences allowing the ‘It is not so that S’ paraphrase. It is plausible to conclude that the contradictory reading will be assigned by the semantics in such cases, unless overridden by a lexical property of the predicate; in the case of shouldn’t and mustn’t, the negative-inflected modal (unlike needn’t, can’t, or doesn’t) is lexically associated with the contrary. If contradictory opposition is not a necessary condition for a negation to be a predicate denial, neither is it sufficient. Each of the sentences in () constitutes a contradictory negation of the corresponding affirmative in (0 ), allowing the ‘it is not so that S’ paraphrase, yet none of them represents a predicate denial or (grammatically defined) sentential negation. () a. b. c. d. e.

It’s impossible for a bachelor to be married. He respects nobody. No bachelors are honest. Not everyone reads Aristotle. Not many children like war.

(0 ) a. b. c. d. e.

It’s possible for a bachelor to be married. He respects somebody. Some bachelors are honest. Everyone reads Aristotle. Many children like war.

Sentence (a) (cf. (b) above) is a clear instance of Aristotle’s term negation, Jespersen’s special negation, or Klima’s constituent negation, as indicated by its form and its behavior with respect to the diagnostics (Klima : –). While (b) displays a mixed pattern with respect to the same diagnostics, the internal position of the negative within the verb phrase once again suggests a constituent negation analysis. Nevertheless, it is clear that a negative in this position can be interpreted with wide scope, although it need not be; cf. Klima (: ) on the ambiguity of I will force you to marry no one. (The availability of wide-scope, contradictory readings for VP-internal negations is further discussed by Bolinger  and Jackendoff ). Sentences (c–e) generally conform to the syntactic diagnostics for S-negation, leading Klima (: ff.) to assimilate such sentences to this category. These examples also constitute logical contradictories of the corresponding positive general statements in (0 c–e). Yet they are not predicate denials. Sentence (c) represents a prima facie instance of constituent or term negation—not predicate term negation as such, but quantifier (subject term) negation, albeit with

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 . 

sentential semantic scope. The negative quantifiers of (d,e) must also involve constituent negation if not everyone, not many children are themselves constituents in these sentences. Indeed, Jespersen (: ) takes (e) to exemplify special negation, where “the negative notion . . . belong[s] logically to one definite idea,” rather than nexal negation, in which negation belongs “to the combination of two ideas,” typically the subject-predicate ‘nexus’. The possibility of paraphrasing not many with few in (d) supports this characterization. (The analysis of negative quantifiers is further investigated in Collins a, .)

.. F   : T    

.................................................................................................................................. In his dictum, “The essence of formal negation is to invest the contrary with the character of the contradictory,” Bosanquet () characterizes the widespread tendency for formal contradictory (wide-scope) negation to be semantically or pragmatically strengthened to a contrary. In this section we explore the contexts in which apparent contradictories are strengthened to receive contrary interpretations. Let us use ©ϕ to represent a contrary of ϕ. Following the theory of opposition outlined above, two contradictories ϕ and ¬ϕ cannot both be false, just as they cannot both be true, while a given proposition and a contrary of that proposition, ϕ and ©ϕ, can both be false, although they cannot both be true. It should be noted that while ¬ is an operator that semantically takes one proposition into another, whatever the grammatical representation of contradictory negation, © is not, since a given proposition may have logically distinct contraries, while this is not the case for contradictories. Geach (: –) makes this point with the example in (). While (a) has two syntactically distinct contradictories, for example Not every cat detests every dog and It’s not every dog that every cat detests, any such co-contradictories of a given proposition will always have the same truth conditions. But (a) allows two contraries with distinct truth conditions, (b) and (c). ()

a. Every cat detests every dog. b. No cat detests every dog. c. There is no dog that every cat detests.

Thus while we can speak of the contradictory of a proposition, Geach observes, we cannot speak of the contrary, but only of a contrary, of a proposition. For our purposes, the crucial logical properties of contrariety are that (i) the contradictory of a proposition ϕ is not a contrary of ϕ and (ii) contrariety unilaterally entails contradiction: ()

a. ©ϕ ⊨ ¬ϕ b. ¬ϕ ⊭ ©ϕ

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



As seen in section ., for Aristotle only sentences can be in contradictory opposition. F and not-F are “logical contraries” excluding a true middle, an existent entity which is neither F nor not-F, although one of the two terms is truthfully predicable of any existent subject in the relevant domain. But naturally occurring cases of prefixal adjectives (e.g. happy/unhappy), including privatives of the kind marked by a(n)- in Greek, may involve an unexcluded middle, as do polar contraries (black/white) or antonym pairs (happy/sad). Modern grammatical discourse departs from Aristotle in allowing for contradictory terms: middle-allowing contrary adjectives (white/black, happy/unhappy) are distinguished from middle-excluding contradictory adjectives (transitive/intransitive, alive/ dead). Against this background, antonyms may be either gradable contraries (happy/sad; tall/short) or complementaries, i.e. gradable contradictories (open/closed; clean/dirty), depending on the presence or absence of a midinterval (Cruse ). (See Lehrer ; Murphy ; and Paradis, Willners, and Jones  for comprehensive overviews on the nature of antonymy.) Jespersen (: ) describes the logical status of negatively prefixed adjectives (see Chapter  in this volume) in English as follows: The modification in sense brought about by the addition of the prefix [un-] is generally that of a simple negative: unworthy = ‘not worthy’, etc. The two terms [F, unF] are thus contradictory terms. But very often the prefix produces a “contrary” term . . . : unjust (and injustice) generally imply the opposite of just (justice); unwise means more than not wise and approaches foolish, unhappy is not far from miserable, etc.

But is this production of the contrary a matter of semantics or pragmatics? Drawing on an epistemic theory of vagueness, Krifka (b) argues that prefixal negation always yields semantic contradictories. On this view, unhappy is literally just ‘not happy’, with the (actual) stronger understanding derived pragmatically. The incomplete cancellation of the two negators in not unhappy is taken to be a purely pragmatic phenomenon, allowing this case to be unified with that of not impossible. But the classical Aristotle/Jespersen theory has compensating advantages of its own. On such an approach, prefixal negatives like unhappy or unkind—like their morphologically simplex classmates sad or cruel—are lexical items that constitute contraries vis-à-vis the corresponding positive. By virtue of their lexical status, they are candidates to undergo further semantic drift, unlike not Adj sequences (or non-Adj forms), as evidenced in the semantic and phonological opacity of infamous or impious. Note too that many un- and iN- adjectives (unkempt, inchoate, incorrigible) lack corresponding simple bases. Furthermore, the prefix non- yields staunch contradictories (typically with objective and/or technical senses) often contrasting minimally with un-Adj or iN-Adj contraries favoring gradable and evaluative contexts: ()

non-American vs. un-American non-Christian vs. un-Christian non-moral vs. immoral non-natural vs. unnatural

non-professional vs. unprofessional non-rational vs. irrational non-realistic vs. unrealistic non-scientific vs. unscientific

Even more problematically for a unified treatment, the treatment of all negatively prefixed adjectives as semantic contradictories of their bases would either extend to simplex

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 . 

antonymic pairs like happy/sad or wise/foolish, where the evidence for semantic contrariety appears incontrovertible, or alternatively posit entirely different semantics for (near) synonyms like unhappy and sad, or unwise and foolish. Thus, while Krifka (b: ) supports the analysis of for example happy and unhappy as “literally contradictories that receive their interpretations as contraries only via pragmatic strengthening,” a traditional (neo-Aristotelian) approach invoking parallel but distinct semantic and pragmatic strengthening routes for contradictory and contrary negative adjectives rests on sounder empirical footing (cf. Horn a). In connection with negatively prefixed adjectives, it is worth mentioning the innovative English “un-noun.” In accordance with the principles governing these productive formations (cf. Horn , b), an un-noun of the form un-X typically refers to an entity that is not an X, but is situated just outside a given category with whose members it shares a salient function. Thus, the un-cola—as popularized in commercials dating back to —is -Up, a carbonated soft drink just external to the cola domain, an un-turkey is a large vegan treat designed for the holiday table, an un-publication is a scholarly chapter that was circulated but never published. Such un-nouns exemplify Aristotle’s privation, a fundamentally pragmatic notion of opposition defined by disappointed expectation: privatives are marked exceptions lacking a property one would expect to find instantiated at the species or genus level. Not just any creature without teeth or eyesight can be called toothless or blind, as Aristotle notes (Categories a–), and so too not just anything that isn’t a cola (e.g. a glass of milk, a hamburger, a T-shirt) would make a good uncola.³ The strengthening of formal contradictory negation (negative element + base) to the status of contrariety as reflected in the semantics of affixal negation (e.g. unfair, dislike) is one symptom of a general pattern or conspiracy we can call MaxContrary (Horn b, a). Another construction illustrating this tendency is neg-raising, the lower-clause understanding of certain higher-clause negations: I don’t think it will rain; She doesn’t want to leave until midnight (see Chapter  in this volume). Whether analyzed as a case of excluded middle (Bartsch ; Gajewski ), an instance of short-circuited implicature (Horn and Bayer ; Horn ), or a syntactic movement rule (Collins and Postal ), the result is a functional contrary in contradictory clothing. An even more straightforward instance of this pattern can be found in the “online” pragmatic strengthening of simple negative statements (I didn’t like the movie). The practice that Bosanquet terms the investment of the contrary with the character of the contradictory can be seen as implicitly invoking the inference schema of disjunctive syllogism or modus tollendo ponens in (): () ϕ ∨ ©ϕ ¬ϕ ∴©ϕ If we accept that Chris is either happy or unhappy, to deny that Chris is happy is to (indirectly) assert that Chris is unhappy. While the key disjunctive middle-excluding ³ As an anonymous referee points out, unX denotes not the contradictory (or immediate contrary) of X—compare non-cola—but an ordinary unexcluded-middle contrary. This supports Aristotle’s reanalysis of privation as the primary contrariety (Metaphysics a).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



premise is typically suppressed, the role of disjunctive syllogism can be detected in a variety of strengthening shifts in natural language where the disjunction in question is pragmatically presupposed in relevant contexts. The availability of strengthened contrary readings for apparent contradictory negation has been recognized since the classical rhetoricians of the fourth century on the figure of litotes, in which an affirmative is indirectly asserted by negating its contrary (Hoffmann ; van der Wouden ). Litotic interpretations tend to be asymmetrical, arising when evaluatively positive but not evaluatively negative: an attribution of ‘not happy’ or ‘not optimistic’ will tend to convey a contrary (in this case ‘rather unhappy’ or ‘fairly pessimistic’), while no analogous virtual contrariety is normally signaled by ‘not sad’ or ‘not pessimistic’, which are usually intended and understood as pure contradictories. This expressive asymmetry is ultimately a social fact motivated by the desire to respect negative face (Ducrot ; Brown and Levinson ; Horn ). In litotes, the interpretation of formal contradictories as contraries arises from the accessibility of the relevant disjunction, triggering the disjunctive syllogism. The homogeneity or all-or-none presupposition (Fodor : ff.) applying to bare plurals, plural definites, and mass predications results in a comparable effect; it is natural to interpret negative statements like Mammals don’t lay eggs, The children aren’t sleeping, or I don’t eat meat as affirmations of contraries rather than understanding them as simple wide-scope negations of the corresponding positives (Mammals lay eggs, The children are sleeping, I eat meat) as would be the case with overtly quantified universals. The relevant principle has been variously formulated: When a kind is denied to have a generic property Pk, then any of its individuals cannot have the corresponding individual-level property Pi. (von Fintel : ) If the predicate P is false for the NP, its negation not-P is true for the NP . . . Whenever a predicate is applied to one of its arguments, it is true or false of the argument as a whole. (Löbner : )

Once again the key step is establishing the relevant disjunction as a pragmatically inferred instance of the law of excluded middle, for example ‘Either mammals lay eggs or mammals don’t lay eggs.’ In fact, this practice was first identified by Aristotle (Soph. Elen. b–a), who offered an early version of the all-or-none (or both-or-neither) in arguing that a negative answer to a “dialectical” or conjoined question like “Are Coriscus and Callias at home?” would imply that neither is at home, given the default supposition that they are either both in or both out. Once again LEM applies where it “shouldn’t”; ϕ ∨ ©ϕ behaves as though it were an instance of ϕ ∨ ¬ϕ, triggering the disjunctive syllogism: ()

(Fa ∧ Fb) ∨ (¬Fa ∧ ¬Fb) ¬(Fa ∧ Fb) (¬Fa ∧ ¬Fb)

The MaxContrary tendency is also illustrated by lexicalization asymmetries and the phenomenon of O > E drift (Horn ), the shift of morpho-syntactic O-vertex quantificational and modal lexical items and collocations (not always, not necessary,

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 . 

possible not) to E meanings (always not, necessary not, impossible). O > E drift is attested cross-linguistically by lexical items like Old English nealles (lit. ‘not+all’) = ‘not at all’, Dutch nimmer (lit., ‘ always’) = ‘never’, or Russian nel’zja (lit. ‘not+must’) = ‘mustn’t’. Crucially, the reverse shift, in which E forms exhibit or develop O meanings, appears to be unattested.⁴ While the traditional logical square in Figure . above is horizontally and vertically symmetrical, natural language lexicalization patterns reveal a sharp asymmetry. Alongside the quantificational determiners all, some, no, we never find an O determiner *nall for ‘not all’/‘some not’; corresponding to the quantificational adverbs always, sometimes, never, we have no *nalways (= ‘not always’, ‘sometimes not’). We find univerbations for both (of them), one (of them), neither (of them), but never for *noth (of them) (= ‘not both’, ‘at least one . . . not’); we have connectives corresponding to and, or, and sometimes nor (= ‘and not’), but never to *nand (= ‘or not’, ‘not . . . and’). Similar if less absolute asymmetries obtain among non-quantificational and indeed non-logical operators; cf. Horn (, ) and van der Auwera (). Various attempts have been made to motivate this asymmetry, including the implicature-based account of Horn (, ) and the alternative approaches in Jaspers () and Seuren and Jaspers (). Several studies have concentrated exclusively on the lexical asymmetries affecting the quantifiers and connectives, excluding modals and adverbials (Huybregts ; Barwise and Cooper ; Hoeksema b; Moeschler , b; Katzir and Singh ; Smessaert and Demey ). Related inquiries have been pursued by Löbner () and van der Auwera (, ) with special reference to the “modal square”, while Ziegeler () proposes typological extensions of the logical geometry. When we leave the quantifiers and connectives behind, we find relative rather than absolute asymmetry—not the fate of the O (un)attested in the paradigms of quantifiers (*nall, *neverybody), connectives (*nor, *noth), and quantificational adverbs (*nalways, *neverywhere) but cases in which a given O form, though attested, can be shown to be less than fully lexicalized, less diachronically stable, more limited in its distribution, and generally marked with respect to its E counterpart. We are dealing here with implicational rather than absolute universals (cf. Greenberg ): essentially, if O then E, but not vice versa. For example, while unlexicalized can not and could not (as in Priests can not marry in section . above) allow both contradictory O (Episcopalian) and contrary E (Catholic) interpretations, lexicalized can’t and couldn’t (along with the orthographic ‘contraction’ cannot and the Scots variant couldnae) can only be used for E statements denoting impossibility. What of A-vertex modals? Whether understood logically, epistemically, or deontically, Eng. must can only incorporate an inner negation, resulting in an E meaning for You mustn’t leave; similarly for You are not to leave. On the other hand, negating the A modal need does yield a lexicalized O value for needn’t, but it can’t help doing so: given the negative polarity status of need as a modal auxiliary—You {needn’t/*need} leave— negation must semantically outscope the modal, rendering the E reading unavailable. (Analogous constraints extend to fellow NPI A-vertex modals Dutch hoeven and Ger. ⁴ The phenomenon illustrated by All that glitters isn’t gold, in which a universal taking apparent wide scope outside negation can semantically scope inside that negation, may appear to be a counterexample, but such “inverse scope” phenomena are actually more complex; for discussion see Horn (: §., §.); Büring (b); de Swart (); Tottie and Neukom ().

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



brauchen; cf. van der Wouden .) It should be noted that needn’t is also distributionally restricted by its semantics (it tends to be deontic) and register (it tends to be constrained to high or written style). In other (spoken and signed) languages, an opaque E-valued modal negation is synchronically unrelated to possibility or necessity, while the corresponding O form for ‘possibly not’/‘not necessary’ is both semantically transparent and non-lexicalized (Horn b). Thus, while the asymmetry in lexicalizing complexes associated with the A, I, and (sometimes) E vertices as against O is exhibited across lexical domains, some domains are more equal than others (cf. van der Auwera  for related discussion). The degree of asymmetry is directly correlated with how closed the category is: strongest for connectives and determiners/quantifiers, somewhat weaker for modal auxiliaries (where needn’t would violate the absolute form of the constraint, and weaker still (though still present) for ordinary predicates. Unlike impossible, unnecessary is restricted to deontic, non-logical contexts and fails to nominalize (impossibility/*unnecessity). (See Horn ,  for elaboration.) The MaxContrary effect is stronger in lexicalized cases than in phrasal complexes. Thus, for example, compare the unambiguous contrary readings required by negative-affixed adjectives like improbable, unlikely, inadvisable with the ambiguity of their unlexicalized counterparts not probably, not likely, not advisable. But while the latter forms can be read as either contradictories or contraries, the contrary E interpretations are more salient, whence the sense that A fair coin is not likely to land heads is false, although this statement is true if read as a contradictory of the affirmative. While the emergence of E readings is particularly robust in the case of incorporated negation, even sequences of ¬ . . . A often strengthen to contrary interpretations. One case in point is the interaction of negation and causatives. Thus, while It. fare + infinitive on its own conveys a strong causative (‘make’, not ‘let’), its negation is often understood with an E-style strengthened meaning: Il caffè non mi fa dormire is understood as asserting that coffee doesn’t let me sleep (rather than that it doesn’t make me sleep). In languages as varied as Japanese, Turkish, Amharic, Czech, Biblical Hebrew, and Jacaltec, the negation of a strong causative (lit., ‘not make’) may or must strengthen to yield contrary (‘make not’ = ‘not let’ = E) force. The reverse drift, in which a ‘not let’ (E) causative is understood as ‘let not’ or ‘not make’ (O), on the other hand, appears to be unattested. A related locus of MaxContrary is the prevalence of prohibitives, as tracked by van der Auwera (a). Roughly two-thirds of van der Auwera’s sample of  languages have a dedicated prohibitive marker, typically derived from the incorporation of a negative element into an imperative or semantically bleached auxiliary. Regardless of the semantic character of the modal or the order of operators, the resultant force is always - (-), i.e. E, not O. There are no dedicated - (-) type O counterparts of E prohibitives. Nor is there even a standard name for such “exemptives.” In addition, some modals are ambiguous or underspecified as between weaker (I) and stronger (A) meanings, this ambiguity disappears under negation. For example, while OE motan could denote permission, ability, or obligation (Goossens : ), its negation ne motan unambiguously signaled E force; comparable facts hold in Dutch (niet moeten) and in non-Indo-European languages. Deal (: ) notes that Nez Perce o’qa ‘must, can’ unambiguously marks impossibility in negative contexts: only the E ‘mustn’t, can’t’, not the O ‘possible not, not necessary’ reading is available.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 . 

.. “L”  

.................................................................................................................................. We observed in section . that for Aristotle, the contradictory negation of a singular (nonquantified) statement arises from predicate denial, a mode of predication by which the predicate applies to the subject, rather than the familiar one-place propositional connective of the kind represented in standard predicate logic. Since it is not a propositional operator, predicate denial cannot apply to a predicate denial, although it can apply to a narrow scope term negation (see Figure .). While Aristotle countenanced multiple negation, to the extent of generating such unlikely sequences as Not-man is not not-just (De Int. b), a given proposition contains at most one instance of negation as wide-scope predicate denial (juxtaposed here with both a negated subject term and a negated predicate term), since each categorical (subject-predicate) statement contains only one predicate. Stoic negation (apophatikon), on the other hand, is an iterating external one-place operator subject to the Law of Double Negation, ϕ ↔ ¬¬ϕ. For Alexander of Aphrodisias (fl.  ), “Not: not: it is day differs from it is day only in manner of speech” (Mates : ). With its propositional connectives and one-place negation operator toggling truth and falsity, the Stoics’ logic of negation prefigured modern (Fregean and Russellian) sentential logic, as well as the precepts of traditional grammar (“Duplex negatio affirmat”). Classical Fregean logic allows for but one negative operator, the contradictory-forming propositional operator applying to a proposition or open sentence, in keeping with “the thesis that all forms of negation are reducible to a suitably placed ‘it is not the case that’ ” (Prior : ). Unsurprisingly, Frege (: ) proclaims the logical superfluity of double negation: “Wrapping up a thought in double negation does not alter its truth value.” Within this metaphor, ¬¬ϕ is simply a way of garbing the thought or proposition ϕ. But even if we admit the Law of Double Negation into our logic, as Hintikka (: ) points out, “in ordinary language a doubly negated expression very seldom, if ever, has the same logical powers as the original unnegated statement.” What exactly is affirmed by double negation? Of course in natural language, multiple negation often fails to affirm or cancel out, given the widespread phenomena of negative concord (I can’t get no satisfaction: see Chapter  in this volume) and expletive or pleonastic negation (I miss not seeing you: see Chapter  in this volume). That is, in ordinary language duplex negatio negat (cf. Horn , a for overviews on multiple negation). But as Jespersen (: ) observes, even when duplex negatio affirmat, the doubly negated property is often not directly affirmed: The two negatives . . . do not exactly cancel one another so that the result [not uncommon, not infrequent] is identical with the simple common, frequent; the longer expression is always weaker: “this is not unknown to me” or “I am not ignorant of this” means ‘I am to some extent aware of it’, etc. The psychological reason for this is that the détour through the two mutually destructive negatives weakens the mental energy of the listener and implies . . . a hesitation which is absent from the blunt, outspoken common or known.

When a negative element is a contrary rather than a contradictory of the corresponding simple affirmative, to deny its application—Chris isn’t unhappy—predictably does not

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



result in mutual annihilation: Chris may be situated in the unexcluded hedonic middle, neither happy nor unhappy but just sort of blaah. But even adjectives that are semantic contradictories, such as impossible, may be coerced under negation into virtual contraries. While it would seem that any action or event must be either possible or impossible, to assess something as not impossible is often to portray its occurrence as a more remote possibility than to assess it as possible simpliciter, as reflected in attestations of ‘It’s possible, or at least not impossible.’ Similar instances of virtual contrariety are readily attested with negated verb phrases or predicate nominals (I didn’t not like it ≠ I liked it) as in ()–(); see Horn (a) for elaboration. () “I didn’t not like it.” “That’s what you call a double negative—do you mean that you liked it?” “I wouldn’t go that far.” (A. M. Homes , May We Be Forgiven, p. ) ()

Rachel wasn’t my girlfriend but she wasn’t not my girlfriend. ( Leslie Gauthier column, )

() Lucy: “Are you friends with Mary?” Robbie: “I’m not not friends with her.”

( episode of tele-drama th Heaven)

Similarly, if we represent the narrow-scope contrariety operator of It is not-white as ©ϕ, its contradictory ¬©ϕ (It isn’t not-white) does not return us to the simple positive ϕ. In fact, the result is an asymmetry reminiscent of that in intuitionistic logic (Heyting ), where the classical law of double negation is valid only for the introduction of consecutive negation operators, as in (a) but not for their cancellation, as in () a. ϕ ⊨ ¬¬ϕ b. ¬¬ ϕ ⊭ ϕ. But while the intuitionists posit just one negation operator, the Aristotelian system distinguishes contradictory (sentential) predicate denial from contrary (constituent) predicate term negation. Thus, apparent contradictory negations, both lexical and syntactic, can be pragmatically strengthened to contraries. Negating such virtual contraries yields the perception of weakened negative force through the recovery of an unexcluded middle. Even when the duplex negatio of ¬¬ϕ seems just to affirm ϕ (e.g. It isn’t impossible), it provides a rhetorically welcome concealment, just as Frege’s metaphor of “wrapping up a thought” in double negation suggests.

A

.................................................................................................................................. I am grateful to Heinrich Wansing and to four anonymous referees for their helpful comments.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  ......................................................................................................................

  Incorporated negation ......................................................................................................................

 

.. I

.................................................................................................................................. N has the formal property to be linguistically marked. The main general way in natural language to mark negation is to use morpho-syntactic devices, that is, expressions which have a precise syntactic configuration. For instance, there is no natural language that can express negative meaning by a particular prosodic contour (Newmeyer ). While negation is encoded in morpho-syntax, it is also encoded in the lexicon, but with a difference in productivity from language to language. This type of negation encoding can be either morphologically marked or incorporated in semantic meaning. For instance, dislike encodes a negative meaning in the prefix dis-, contrasting with the positive predicate like, whereas the verbal predicate hate does not signal, by a morphological marking process (derivation), its negative content. As these different ways in signalizing negative content show, the first issue is to disentangle the different formal ways of expressing a negative meaning from the variation in meaning between these cues. A second issue is the nature of semantic and pragmatic relations between incorporated negative meaning and its positive counterpart. One classic explanation (Horn , b) is to make a strong distinction between two types of semantic negative operator: a contradictory negation (¬), which is truth-conditional and inverts the truth-conditional meaning of the proposition, and a contrary negation (©), whose main semantic property is to restrict the assertability of contrary predicates: for instance, like and dislike, as well as love and hate, cannot be true together, whereas they can be false together. This property, which is logically predicted by the logical relations between antonyms (see Figure .), is not the case with contradictory negation, that is, the morpho-syntactic negative marker (English not, French ne . . . pas, etc.). In case of a contradictory negation, only one of the alternatives is true, either the positive or the negative proposition. Third, the lexicalization of negative meaning gives rise to asymmetrical inferences. For instance, in some predicates such as antonyms, like nice vs. nasty, the negation of the negative predicate does not give rise to the same inference with the negation of the positive

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 



one—French pas méchant ‘not nasty’ is compatible with gentil ‘nice’, whereas pas gentil ‘not nice’ means méchant ‘nasty’. In parallel, it has been observed that the negation of a negative predicate gives rise to what is known under the label litotes, or meiosis. One famous example is Chimène’s exhortation to his lover Rodrigue, Va, je ne te hais point (‘Go, I do not disdain you’), mandatorily implicating ‘I love you’ (Pierre Corneille, Le Cid). The crucial point is to explain why such an implicature is not mandatory when the positive predicate is under negation: I don’t love you has only as one of its possible meanings the strongest one, that is its negative counterpart (‘I hate you’), even though this meaning, unless it is contextually encouraged, is not its preferred reading. This asymmetry will be explained by considering the double negative meanings, both the contradictory and the contrariety negations. Finally, it has been observed that incorporated negation is incompatible with metalinguistic negation, that is, non-truth-conditional meaning. The main question is thus the following: why is metalinguistic interpretation banned in incorporated negation? The challenge of this issue is to present a consistent perspective including semantic and pragmatic types of meaning. Some recent findings on metalinguistic negation (Moeschler ) will be recalled, giving a consistent picture of what the meaning of negation is. The main difference is located in the type of inference that both descriptive and metalinguistic uses of negation give rise to, contrasting with the entailments and the implicatures of incorporated negation. This chapter is organized as follows. Section . discusses the relation between types of negation and types of negative meanings, and also addresses the issue of incorporated negation. Section . discusses the issue of the relation between lexical categories and negative predicates. Section . introduces the two Hornian negation operators, ¬ (contradiction) and © (contrariety), and argues that negative predicates trigger an implicature, whose meaning corresponds to the negation of its positive counterpart. Section . capitalizes on this distinction to explain the pragmatic effects of syntactically negated negative predicates: on the one hand, data show conventionalization in meaning (with evaluative predicates) and, on the other hand, frequent uses of litotes implicate their positive counterpart predicates. Finally, section . addresses the issue of the absence of metalinguistic negation in negative predicates. The conclusion (.) summarizes the six semantic and pragmatic properties of negative predicates.

.. T       

.................................................................................................................................. Incorporated negation consists in encoded negative meaning within lexical items. In other words, negation can be signalized either overtly by morphological marking (as negative prefixes in Romance and Germanic languages) or covertly because of the negative meaning of the lexical unit. As far as incorporated negation is concerned, the contrast with morpho-syntactic devices of negation, as English not, French ne . . . pas, is not only a difference in the pattern in which

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

negative meaning is linguistically encoded (syntax vs. lexicon, ..), but also a difference in types of meaning (..).

... Types of negative devices First of all, what makes the main difference between syntactic negation and lexical or incorporated negation, either overtly or covertly, is that the syntactic negation is ambiguous.¹ In a simple syntactic negation as (), at least two (descriptive) meanings are available (), whereas its lexical negative counterpart () is not ambiguous: ()

Abi is not married.

()

a. Abi is not-married b. It is not the case that Abi is married

()

Abi is unmarried.

In (), either negation scopes over the predicate married, which implies that another property is true of Abi, for instance ‘single’ or ‘divorced’, whereas (b) simply states that the proposition ‘Abi is married’ is false of Abi. On the other hand, () denies that Abi has the property of being married, since unmarried entails ‘not-married’. As a consequence, the difference between () and () is not a difference in truth conditions, in the descriptive meaning of (). Another semantic property, completeness, will be introduced later (.). The second main difference between () and () is that only syntactic negation can take a semantic wide scope, that is, have a metalinguistic usage, which is not possible for incorporated negation. For instance, () can receive two readings, as () shows, whereas () cannot receive a metalinguistic reading, as () shows (Moeschler ): ()

Abi is not happy.

()

a. Abi is not happy, she is less than happy. b. Abi is not happy, she is ecstatic.

()

Abi is unhappy.

()

a. Abi is unhappy, she is less than happy. b. #Abi is unhappy, she is ecstatic.

Third, there is a strange phenomenon regarding negative quantifiers. Whereas negative universals are lexicalized, as no or none, there is no negative corresponding word for some, that is, no natural language expresses a lexicalized negative particular quantifier, meaning ¹ This chapter will not develop the difference between constituent negation and sentential negation, as first proposed by Klima (), mainly because the contrast developed here is between syntactic sentential negation and incorporated negation. See Chapter .

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 



‘not all’ (Horn ; Moeschler a): for instance, English does not have as a quantifier a word like nall, meaning ‘not all’. So, negative quantifiers are restricted to no (none) and a syntactic relation, combining a positive particular and negation: not all or some not, exemplified in (a–b), as opposed to the non-attested lexicalized negation in (c): ()

a. Not all students passed their exam. b. Some students did not pass their exam. c. *Nall students passed their exam.

These differences in meaning must be explained. One of the main goals of this chapter is to make explicit the difference in meaning between syntactic negation and incorporated negation. The second purpose of this chapter is to explain why incorporated negation cannot have a metalinguistic usage (cf. .).

... Types of negative meanings The second difference to be made explicit is the different types of negative meanings (Chapter ). The main two differences are first a scope difference and second the relation between the logical meaning of negation and its meaning in use. As has been mentioned before, negation can take narrow (restricted to the predicate) or wide scope (having the very act of utterance as scope), giving rise, respectively, to descriptive and metalinguistic negation. For instance, negation takes narrow scope in its descriptive meaning (a), as the classic king of France example shows (Russell  for the introduction of this example; Strawson  for a pragmatic counter-analysis), as well as wide scope in its metalinguistic reading (b): ()

a. The king of France is not bald. b. The king of France is not bald; there is no king of France.

() and () make these readings explicit: narrow scope is expressed by an internal negation, scoping over the propositional function (x), whereas wide scope corresponds to external negation, scoping over the whole proposition () (see Horn ): ()

9x [(x) ∧ ¬9y [ (y) ∧ (x ≠ y)] ∧ ¬  (x)] ‘there is an x such that x is a king and there is no y which is a king and different from x and x is not bald’

()

¬9x [ (x) ∧ ¬9y [ (y) ∧ (x ≠ y)] ∧  (x)] ‘there is no x such that x is a king and there is no y which is a king and different from x and x is bald’²

² The assumption is that wide scope of negation (¬9) equals metalinguistic negation, that is makes the presupposition of existence false. On the other hand, narrow scope (9¬) preserves the existential presupposition. When negation has wide scope on scalar predicates (see section .), wide scope targets the scalar implicature of the positive counterpart (Moeschler , ).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

In other words, what () makes explicit is that the property of being  is false of the king of France, whose existence is taken for granted. On the other hand, what is made explicit in () is that the proposition ‘the king of France is bald’ is false because of the absence of such an individual described as the king of France. Contrastively, incorporated negation cannot have wide scope: for instance, (a), which expresses the contrary meaning of (a), cannot be associated with any readings compatible with the non-existence of the king of France. ()

a. The king of France is haired. b. # The king of France is haired; there is no king of France.

In a nutshell, what can be expressed by incorporated negation cannot be wide scope, but only narrow scope. The second issue, as regards types of meaning, is the classic difference, made explicit in Horn (b), between contradictory and contrary negation. This difference will be made explicit in section ., but, for now, what is important to stress is that syntactic negation can be put in correspondence with logical negation, whose meaning is to reverse the truth value of the proposition expressed. Logical negation is a one-argument function, whose meaning is, in bivalent logic, to yield for a given proposition the proposition with the opposite truth value, as Table . shows.

Table .. Truth table for logical negation P  

¬P  

This logical meaning explains why descriptive interpretation of negation in (a)—the king of France is not bald—makes the logical form (), as well (), true: in (), the proposition function  (x) is evaluated as false, and thus ¬  (x) is true, as well the other propositional functions. As () is a conjunction of propositional functions, it is true only if all such propositional functions are true. In (), the whole proposition is true if what is under the scope of negation is a false proposition: as the propositional function  (x) is false, because its contradictory proposition is entailed by the corrective sentence there is no king of France, negation scoping over a false proposition yields a true proposition. However, it has been argued in Horn (, ) that metalinguistic negation is nontruth-conditional, that is, is not the result of a truth-conditional operator of negation. Horn proposes to analyze metalinguistic negation in a speech act way: metalinguistic negation corresponds to the operator ‘I object that U’, where I object that is the metalinguistic negation scoping over the utterance U. So, (b) receives as a paraphrase (): as a performative formula, it cannot be true or false, mainly because ‘I object’ has no truth conditions, but uses conditions, defined in the standard speech act analysis by semantic rules of assertive speech acts (Searle , ):

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  ()



I object that the king of France is bald; (because) there is no king of France.³

What about the nature of lexically incorporated negation? As will be argued later, the negative component cannot be reduced to logical negation or to the metalinguistic negator. Let us examine this issue with a discussion on negative causative verbs. The second issue (the incompatibility between incorporated and metalinguistic negation) will be discussed in section ..

... What makes negative meaning in incorporated negation? What is striking with incorporated negation is that a negative component is incorporated in the semantic meaning of the predicate. The main question is where such a negative meaning is encoded. A first hypothesis is that such a negative component cannot be located in the entailment of the sentence, because it would give a negative meaning to a positive predicate. For instance, kill is a two-place predicate whose entailment clearly contains a negative operator, as the meaning postulate () shows (see Gordon and Lakoff  for the concept of meaning postulate): ()

∀x∀y [ (x,y) →  (x,  (¬ (y))] ‘for every x and for every y, if x kills y, then x causes y to become not alive’

The crucial point in this analysis is the occurrence of a negative operator (¬). So, there is, in the entailment of the predicate kill, a negation, expressed by ¬(y). The question is therefore the following: is kill a negative predicate? This question is crucial because an entailment is defined as not defeasible, as opposed to a conversational implicature, which is cancellable (Grice ; Sadock ). It is also crucial for scalar implicatures (Gazdar a; Horn , ; Levinson ; Zufferey, Moeschler, and Reboul ) because they are negative too (see Levinson ), as () shows: the content of the scalar implicature contains a negation, which cancels the stronger term (all for some): ()

a. Some of the students passed their exam. b. Not all of the students passed their exam.

The same reasoning can be made for conventional implicatures and lexical (semantic) presuppositions, too. For instance, only conveys a conventional implicature (CI) ‘no more than’, as in (). Similarly, aspectual verbs like begin have a negative lexical presupposition (LP), as in (): ()

Abi is only beautiful. CI: Abi is no more than beautiful.

³ In Strawson’s analysis of the king of France example (Strawson ), the falsehood of the existential presupposition makes the utterance meaningless, that is, neither true nor false. Presuppositions are thus conditions of uses for utterances.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

()

Abi began smoking. LP: Abi was not smoking before.

So, the question is whether these negative inferences (entailments, scalar implicatures, conventional implicatures, and presuppositions) make their lexical triggers negative. A positive answer would give rise to strong difficulties for defining what a negative predicate is, because it would imply that both semantic encoding and pragmatic encoding belong to the conceptual meaning of a lexical item.⁴ However, other conventional meanings give rise to the same issue, even if they are clearly semantic. Take the example of antonyms. In antonymy, the semantic relation is straightforward: if A and B are antonyms, then A entails not-B and B entails not-A, whereas not-A does not entail B nor not-B entails A (Ducrot ). So, the negative feature of a predicate cannot be located in the negative polarity of the entailment. If this were the case, then both antonyms would be negative predicates. For instance, both beautiful and ugly should be defined as negative, because of their negative entailments, as made explicit in () and (): ()

Abi is beautiful → Abi is not ugly

()

Abi is ugly → Abi is not beautiful

This is however highly controversial, because only ugly can be defined as a negative predicate. So, a predicate is not negative because its entailment is negative, but because of the polarity of its inferred meaning. In other words, whereas beautiful is positively oriented—it can be used as an argument for a positive conclusion—, ugly is negative, because it can only support negative conclusions (Ducrot ). The types of discourse sequences in () and () show that argumentative polarity crossovers are impossible: ()

a. Abi is beautiful: she can play Juliet. b. # Abi is beautiful: she cannot play Juliet.

()

a. Abi is ugly: she cannot play Juliet. b. # Abi is ugly: she can play Juliet.

Another example of a negative predicate can be given by negative causatives. Take the example of the French causative empêcher ‘prevent’, as illustrated in () and (): ()

Jacques a empêché les enfants de manger le gâteau. Jacques has prevented the children to eat the cake ‘Jacques prevented the children from eating the cake.’

()

Jacques n’a pas empêché les enfants de manger le gâteau. Jacques  has  prevented the children to eat the cake ‘Jacques did not prevent the children from eating the cake.’

⁴ For the conceptual-procedural distinction, see Blakemore (), Moeschler (), Wilson (), Wilson (), Wilson and Sperber (: ch. ).

 



In (), the entailment ‘the children ate the cake’ (e2) is true, as represented in Figure ., in history h0 , whereas it is false in history h, since e1—the event of Jacques’s preventing the children from eating the cake in ()—blocks e2. Reboul (: ) gives the following representation in a ramified-time branching approach (Xu ) shown in Figure .. h

h’ e2

h’’ e3

e1

m

 .. A time-branching representation for negative causatives (from Reboul : )

In other words, m is the moment where alternatives of events are possible (this is a transition): e3 in history h00 is the event ‘the children ate the cake’, where there is no obstacle preventing the occurrence of the event, illustrated for instance in (): ()

Jacques a laissé les enfants manger le gâteau. Jacques has let the children eat the cake ‘Jacques let the children eat the cake.’

e1 is the preventing event, and in history h (), there is no occurrence of the eating event. History h0 gives rise to the event e2 (the children ate the cake), where the preventing event fails. In other words, the preventing event blocks the occurrence of the event e2, explaining its entailment, as in (): ()

a. Jacques a empêché les enfants de manger le gâteau. Jacques has prevented the children to eat the cake ‘Jacques prevented the children from eating the cake.’ b. Entailment: the children did not eat the cake

So, empêcher would receive the following semantic description: ()

Lexical entry for empêcher ‘prevent’: x empêcher e ¼df x  ¬e⁵

As a result, the first positive conclusion of this chapter is the following: ()

What defines a negative predicate is not a negative entailment, but a negative feature encoded in lexical meaning.

⁵ () encodes negation not as the result of an entailment, mainly because the  predicate is not a logical consequence but a part of the lexical meaning of empêcher ‘prevent’.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

.. C    

.................................................................................................................................. In section ., a preliminary definition of what a negative predicate is has been given, without defining the lexical domain of negative predicates. The first hypothesis is about the possible grammatical categories of a predicate. Lexical categories, like nouns, verbs, and adjectives, are natural candidates, because they all behave semantically as predicates.

... Denotation of predicates Semantically, a predicate is a function having individuals as arguments. For instance a oneplace predicate is a function from a set of individuals into a truth value, or put in a simpler way, the denotation of a one-place predicate is a set of individuals; a two-place predicate is a function from a set of individuals into a one-place predicate, that is, a function from a set of individuals into a truth value; and so forth for three-place predicates, etc. Intransitive verbs, some common nouns, and adjectives are examples of one-place predicates. So, such a semantics predicts that negative predicates are encoded in common nouns, verbs, and adjectives, because they are one-place predicates (adjective and common nous) or one/two/three-place predicates (verbs). () to () are some examples of such negative predicates: ()

Nouns: hate, destruction, ugliness, nastiness . . .

()

Adjectives: ugly, nasty, unhappy . . .

()

Verbs: hate, dislike, destroy, prevent, hinder . . .

The reason why these predicates are negative is the same: their lexical content encodes a negative feature. However, this feature does not license the same set of entailments as syntactic negation does. As will be developed in section ., there is no contradiction between hate and love, ugly and beautiful, like and dislike. So, the effect of the negative feature is not a logical negation scoping over a proposition. In other words, if a negative predicate is false of its argument, this does not imply that its positive counterpart is true: the negation of a negative predicate can be compatible with the negation of its positive counterpart. In other words, whereas (a) and (b) cannot be true together, (a) and (b) can be true together: ()

a. Abi is beautiful. b. Abi is not beautiful.

()

a. Abi is not ugly. b. Abi is not beautiful.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 



Given that incorporated negation is encoded within negative predicates corresponding to lexical categories, the question is whether negative predicates can be encoded in functional categories, as prepositions, quantifiers, or connectives. Some facts argue for a negative answer. First, there are negative connectives: concessive connectives (although, unless), as well as contrastive connectives (but), encode a negative meaning. Can they be considered negative predicates? If discourse connectives are defined as two-place predicates, can we assume that English but, French mais, German aber encode a negative feature? In fact, the comparison between positive and negative connectives does not allow drawing such a conclusion: but has the same truth conditions as and, if we assume a classic Gricean perspective (Grice ), summarized in Table ., but its pragmatics constrains the research of a contrast at the implicature level between the connected discourse segments. In a classic example of but (Lakoff ), the first discourse segment implicates the negation of the second (): ()

He is a Republican, but he is honest. Implicature: he is a Republican +> he is not honest

Table .. Truth table for the conjunction P    

Q    

P∧Q    

Second, some prepositions are negative, as without, contrasting with with. However, in does not have a negative counterpart; the same holds for by, for, of, whereas spatial prepositions can be in a contrast relation: up vs. down, on vs. under, near vs. far. However, in this case, the contrast does not entail a negative meaning: for instance, down indicates a level of height, and not a contrary relation to up; far makes explicit a long distance, and is not the contrary to near, etc.⁶ Third, negative determiners and quantifiers are lexically realized. However, in French for instance, there is no negative counterpart to quelques or certains ‘some’, whereas aucun ‘no’ is contrary to tous ‘all’. In a nutshell, negative predicates are limited to conceptual meanings, encoded in lexical categories.⁷ The explanation proposed is semantic and is made explicit in ():

⁶ The reason is because these predicates can be in correspondence with measures in a same scale. See Ducrot () for such an implementation for gradable adjectives. ⁷ For a more complex description of lexical categories, see Moeschler () and Moeschler ().

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

()

Only lexical categories having as denotation sets of entities or sets of sets of entities can be negative.⁸

In brief, functional categories are excluded for the following reasons. Connectives have as denotations relations between truth values, prepositions are relations between entities, and quantifiers and determiners have as denotation relations between predicates, that is, functions from sets of individuals to truth values. There seems to be, therefore, a semantic constraint on negative predicates: they all must be about sets of entities, that is, describe sets of entities that are excluded from the denotation of the predicate.

... A first explanation How is it possible to make sense of this finding? Why are negative predicates restricted to lexical categories, and not to functional ones? By restricting negative predicates to categories whose denotations are sets of entities, and not to functional categories, I would like to make sense of what the semantics of predicates is. Let us first compare negative descriptive meaning expressed by syntactic negation, as in (): ()

Abi is not married.

What () means is that the individual Abi is defined by a set of properties excluding the property of being . In other words, in order to get a complete interpretation of (), what is required is to find the appropriate predicate true of Abi which belongs to the set of alternatives to . Such a predicate can be given by a corrective clause, as in (): ()

Abi is not married but engaged.

So, the corrective clause is relevant inasmuch it makes some possible alternatives true, as in (): ()

Abi is not married but {single, divorced, a widow}.

The contrast with a negative predicate is now more than relevant. By saying that Abi is unmarried, what is at stake is not a set of alternatives to : the negative predicate is sufficiently relevant because of its entailment—if Abi is unmarried, she is not married. But if Abi is not married, she can be more than unmarried: for instance, single or engaged. ()

Abi is unmarried → Abi is not married

⁸ Notice that this definition does not apply to quantifiers, because their semantics are  between sets of individuals (see Barwise and Cooper ).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 



Moreover, the search for an alternative to a negative predicate is odd: first, it cannot be a corrective clause (), and second, an explanation clause gives a metarepresentation meaning of the negative predicate (unmarried), that is, makes explicit the reason why this proposition has been asserted (): ()

# Abi is unmarried but engaged.

()

Abi is unmarried, because she is engaged.

So, whereas syntactic descriptive negation implies (in a non-technical sense) an incomplete pragmatic interpretation, the use of an incorporated lexical negation gives rise to a set of possible implicatures, not to a set of possible alternatives. These implicatures are not generalized, but particularized: for instance, regarding Abi is unmarried, ‘she is looking for a serious relationship’, ‘it can be your chance’, etc. The meaning completeness of negative predicates is thus the third provisional conclusion of this chapter. So, in this section, three major semantic conclusions, albeit provisional, have been proposed: ()

Semantic conclusions a. A negative predicate encodes in its lexical meaning a negative feature; b. only lexical categories having as denotation sets of entities or sets of sets of entities are candidates for negative predicates; c. negative predicates are semantically complete.

Now, what is the semantic property of the negative feature defining a negative predicate?

.. C     

.................................................................................................................................. What is crucial is to define the type of meaning relation implied by negative predicates. Let us take as a first example a morphologically unmarked negative predicate: ugly is the contrary predicate of beautiful, which implies that an entity x cannot be both beautiful and ugly, whereas their contradictory counterparts can be true together. In other words, in a pair of antonyms , the following semantic meaning relations are the case: () If A and B are contraries, then a. A entails ¬B b. B entails ¬A c. ¬A and ¬B are subcontraries. In the case of beautiful and ugly, these relations can be made explicit by the logical square in Figure ..

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



  positive predicate beautiful

negative predicate contraries

ugly

contradictories

not ugly

not beautiful subcontraries

 .. Meaning relations for contraries

This square of relations provides a first answer concerning the nature of the negative feature defining a negative predicate: as concluded in section ., this property cannot be defined in terms of negative entailment, because it symmetrically applies to positive predicates as well. So, one possible answer is that the negative feature encoded in a negative predicate is the contrary negation ©: a positive predicate P entails ¬©P (beautiful entails ‘not ugly’), whereas a negative predicate ©P entails ¬P (ugly entails ‘not beautiful’). So, what makes a predicate negative is the presence of © in its lexical meaning. These relations strongly contrast with scalar predicates. In this case, the semantic relation between predicates is distributed in scales, each predicate belonging to a positive (no negation) or to a negative scale, containing negated predicates. Figures . and . are illustrations for positive and negative scalar predicates. positive scale gorgeous

negative scale contraries

not beautiful

contradictories

beautiful

not gorgeous subcontraries

 .. Meaning relations for positive scalar predicates positive scale hideous

negative scale contraries

not ugly

contradictories

ugly

not hideous subcontraries

 .. Meaning relations for negative scalar predicates

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 



What then are the differences between antonym predicates and scalar ones? First, if we add to logical relations pragmatic meaning relations as conversational implicatures, we obtain different inferences, shown in the comparison between Figure . and Figures .–.. In the latter situation, whatever the polarity of the predicate is, and are semantic scales: the strong predicate entails the weak one, and the weak predicate implicates the negation of the strong one (Horn ; Gazdar a; Levinson ). So, by saying Abi is ugly, the speaker implicates that she is no more than ugly, that is, she is ‘not hideous’, by reference to the first Gricean quantity maxim (“Make your contribution as informative as is required (for the current purpose of the exchange),” Grice : ). Exactly for the same reasons, Abi is beautiful implicates that she is no more than beautiful, that is, ‘not gorgeous’. Second, the crucial question is what happens when the negation of a positive or negative antonym occurs, that is, within the semantics ¬P and ¬©P? Antonyms do not function as scalar predicates. First, they are not scalar, in the sense that there is no positive entailment relation between them (beautiful entails ‘not ugly’, whereas gorgeous entails ‘beautiful’). Second, antonyms belong to different scales (positive and negative) and when they are negated, they invert the argumentative orientation. For instance, not beautiful will support a negative conclusion (Ducrot ). This property can be explained by the logical square in Figure .. As represented, not ugly and not beautiful are lower-bound and their upper-bound correlates are not negated like scalar predicates are: for instance, not ugly does not implicate not beautiful, as beautiful implicates not gorgeous, or some implicates some not (Horn ). On the contrary, not beautiful may implicate ugly, even though this implicature is not mandatory. This can be shown by the even test, making the lower-bound implicatum not contradictory (), and with the logical compatibility between the two subcontraries (): ()

a. Abi is not beautiful, she is even ugly. b. Abi is not ugly, she is even beautiful.

()

Abi is neither beautiful nor ugly; she is just ordinary.

Hence, the implicature triggered by ¬P and ¬©P are not scalar: in the neo-Gricean literature, they are labeled as I(nformative)-implicatures (Levinson ) or R(relation)implicatures (Horn ). Such implicatures are the result of the principle of informativity and the R-principle, stating respectively “Say as little as necessary” (maxim of minimization, Levinson : ) and “Make your contribution necessary ( . . . ). Say no more than you must” (Horn : ). Following Horn (), the R-principle is an upper-bounding principle, inducing lower-bounding implicata. In Horn (), a different interpretation is given, explaining the implicature as the result of the MaxContrary effect. This effect, which is somehow parallel to the effect of Neg-Raising (Horn ),⁹ gives a lower strength to syntactic negation, as the scale shows. This is also the case for , , , etc. The even test in () shows indeed that a negative predicate, morphologically overt or covert, is a stronger negative statement than a syntactic negation. We face here a first, pragmatic, conclusion, complementary to the three semantic conclusions (): ()

First pragmatic conclusion on negative predicates: Negative predicates (lexically incorporated negation) are stronger negative statements than syntactic negative utterances (syntactic negation).

So, from a pragmatic perspective, the existence of negative predicates, compared to syntactic negation, is mainly to avoid that the I/R-implicature can be defeated. Because ugly is stronger than not beautiful, the use of a negative predicate makes the assertion unambiguous: the question of whether the implicature of the negative statement has to be drawn or not does not arise, because the negative predicate, that is, the negative feature of the predicate (©P), is not cancellable. This is good news: negative predicates are motivated semantically as well as pragmatically.

.. S     

.................................................................................................................................. Still, there is a puzzle in the behavior of antonyms: with some subjective antonym predicates, there is an asymmetry between the negation of the negative predicate and the negation of its positive counterpart. For instance, in French, pas gentil ‘not kind’ is used to convey the negative antonym meaning (©P), that is, méchant ‘nasty’, whereas the negation of the negative predicate pas méchant ‘not nasty’ does not give rise to the entailment gentil ‘kind’(): ()

Pierre n’est pas gentil: il est au contraire méchant. Peter  is  nice: he is on the contrary nasty ‘Peter is not nice: on the contrary, he is nasty.’

()

?? Pierre n’est pas méchant: il est au contraire gentil. Peter  is  nasty: he is on the contrary nice ‘Peter is not nasty: on the contrary, he is nice.’

In other words, the implicature from ¬P to ©P is quasi-mandatory, shown by (): ()

# Pierre n’est pas gentil ni méchant non plus. Peter  is  nice nor nasty either ‘Peter is neither nice nor nasty either.’

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 



On the contrary, with a negated negative predicate, as pas méchant ‘not nasty’, the lowerbound implicature is not required: pas méchant is positively oriented but targets an intermediate state between kindness and nastiness. However, this is problematic, because it implies a scale from kindness to nastiness, where pas méchant ‘not nasty’ targets an intermediary state, as shown in (): ()

gentil > pas méchant > méchant ‘nice > not nasty > nasty’

However, such a scale does not correspond to any semantic structure. First, the reverse scale () has no empirical justification: ()

méchant > pas gentil > gentil ‘nasty > not nice > nice’

Second, how to locate the negation of the predicate, respectively pas gentil ‘not kind’ in () and pas méchant ‘not nasty’ in ()? Is there a quantitative difference in kindness or in nastiness? All linguistic tests, for instance the even test, give a negative result, since both () and () are acceptable, proving no quantitative difference between pas gentil ‘not kind’ and pas méchant ‘not nasty’: ()

Pierre n’est pas gentil, ni même pas méchant. Peter  is  nice nor even  nasty ‘Peter is neither nice nor even nasty.’

()

Pierre n’est pas méchant, ni même pas gentil. Peter  is  nasty nor even  nice ‘Peter is neither nasty nor even nice.’

So, another explanation must be given. Before attempting to formulate such an explanation, I would like to show a parallel asymmetry, with the famous reply from Chimène to Rodrigue, after he had claimed that his love for Chimène is impossible.¹⁰ ()

Va, je ne te hais point. go I  you hate  ‘Go, I do not disdain you.’

¹⁰ “Rodrigue: Au nom d’un père mort, ou de notre amitié, Punis-moi par vengeance, ou du moins par pitié. Ton malheureux amant aura bien moins de peine À mourir par ta main qu’à vivre avec ta haine. Chimène: Va, je ne te hais point.” (Corneille, Le Cid, Act III, Scene ) “Rodrigue: In the name of a dead father, or our amity, Punish by vengeance, or at least by pity. Your unfortunate lover finds here less pain, Death at your hand, than life with your disdain. Chimène: Go, I do not disdain you.” (Translated by A. S. Kline, The New York Public Library: Digital Collections, , consulted on August , )

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

This is a famous example that rhetoric calls litotes, or meiosis, that is, saying the less to implicate the more. Clearly, by saying she does not disdain Rodrigue, Chimène strongly implicates that she loves him. What is striking is that this effect only occurs with a negative predicate. Imagine that, in an alternative Le Cid, Chimène says to Rodrigue on the contrary (): ()

Va, je ne t’aime pas. go I  you love  ‘Go, I do not love you.’

Is the stronger interpretation ‘I hate you’ expected? This is not obvious, because the more expected formulation, at least in colloquial French, would be is something like (): ()

Dégage, je te déteste. go away I you hate ‘Go away, I hate you.’

So, we face here two contradictory conclusions: first, pas méchant ‘not nasty’ blocks the lower-bound implicature, whereas in the litotes case, the negation of the negative predicate mandatorily implicates the positive one. How can we explain this contradiction? In the case of pas méchant ‘not nasty’, a possible explanation is that this expression is in a process of conventionalization, what Morgan () calls a convention of use. The frequent use of pas méchant ‘not nasty’ as meaning a positive but weaker degree of kindness than gentil ‘kind’ can explain that the implicature gentil ‘kind’ cannot be obtained. However, what is striking is that the polarity changes: pas méchant ‘not nasty’ is positively oriented. As being formally negative, it is weaker than any other positive predicate. Last but not least, this process is not unproductive lexically: it happens mainly with evaluative adjectives (non-classifying predicates in Milner ), as c’est pas bête ‘it is not stupid’, c’est pas con ‘it is not stupid’, c’est pas grave ‘it is not serious’, whereas the negation of a positive evaluative predicate, as in c’est pas intelligent ‘it is not clever’, clearly implicates its contrary ‘it is stupid’. In the case of litotes (je ne te hais point ‘I do not disdain you’), the mandatory implicature is mainly the result of an impossible explicit love declaration. Chimène cannot declare to Rodrigue je t’aime ‘I love you’, because of the conflict between Chimène and Rodrigue’s families, and also because the concept  has not been put on the table: the question under discussion (QUD) (Roberts ) is whether Chimène disdains Rodrigue or not. He himself introduces the concept  ‘’: Ton malheureux amant aura bien moins de peine À mourir par ta main qu’à vivre avec ta haine (‘Your unfortunate lover finds here less pain, Death at your hand, than life with your disdain’). In other words, the concept  is introduced by Rodrigue’s last words, and Chimène takes the opportunity to link up her love declaration with the negation of the negative predicate ï ‘’. So, there is both a stylistic and a communicative reason why a negative predicate is used under syntactic negation. Does it mean that the use of a negated negative predicate with a mandatory implicature is a non-ordinary use of language, that is, intending specific stylistic effects? On the

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 



contrary, it is hypothesized here that these effects are just the result of semantic and pragmatic properties of negative predicates. They are particularly relevant with psychological and subjective predicates, because negated negative predicates are weak forms of strong content. This psychological reflex is particularly well designed in a Swiss French expression, déçu en bien ‘positively disappointed’, which combines a negative predicate (déçu ‘disappointed’) and a positive adverbial (en bien ‘positively’). The déçu en bien implicature is something as ‘positively surprised while having negative expectations’. So, our second pragmatic conclusion is the following: ()

Second pragmatic conclusion: Negative predicates are used in the scope of syntactic negation to mitigate the alternative explicitness of a positive, conversationally implicated, predicate.

.. W       ?

.................................................................................................................................. The last issue to be discussed is the incompatibility between negative predicates and metalinguistic negation (see Chapter  on metalinguistic negation). It has been observed, since Horn (), that metalinguistic negation is not possible with negative predicates () nor with negative polarity items (NPIs) () (see Chapter  in this volume): ()

a. * It is impossible for you to leave, it is necessary. b. It is not possible for you to leave, it is necessary.

() a. * Chris didn’t manage to solve any of the problems—he managed to solve all of them. b. Chris didn’t manage to solve some of the problems—he managed to solve all of them. It is also noticeable that with covert negative predicates, there is no possible use of metalinguistic negation: whereas I don’t love you can be interpreted metalinguistically (a), it is not the case for the negative incorporated predicates (b): ()

a. I don’t love you, I adore you. b. # I hate you, I adore you.

Conversely, a negative predicate can also be under the scope of metalinguistic negation as well (a), and, similarly to (b), a positive predicate cannot be used with its implicature meaning in metalinguistic context (b): ()

a. I don’t hate you, I detest you. b. # I love you, I detest you.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

The question of the impossibility of a metalinguistic use with a negative predicate raises the issue of the nature of the negative feature encoded in negative predicates. It has been assumed that this feature corresponds to ©, the contrary negation. So, another way to address the issue of the non-metalinguistic usage of negative predicates is to ask why © cannot be a metalinguistic operator. In other words, why is ©P not compatible with the meaning ‘I object that P’? For instance, why does (a) not mean (b)? ()

a. Abi is unmarried. b. I object that Abi is married.

The possible answers to this question are multiple. First, metalinguistic usages of negation require, though not mandatorily, a corrective clause, which is not the case with a negative predicate because of its completeness. Second, the corrective clause (COR) entails, at least for metalinguistic uses of upward negation (Moeschler , ), the positive clause (POS). For instance, (a) triggers entailment (b) (see Figure .). As (a) has no COR, no entailment leading to POS is available. ()

a. Abi is not beautiful, she is gorgeous. b. Abi is gorgeous → Abi is beautiful

Third, metalinguistic negation is metarepresentational: it means that by objecting the very utterance, the speaker intends to mean that she cannot affirm POS. This is particularly well demonstrated when negation scopes on the form vs. the content of utterance. () are classical examples of metalinguistic negation on the form (Horn ; Wilson ; Carston ; Wilson and Sperber : ch. ): ()

a. Around here we don’t eat tom[eiDouz] and we don’t get stressed out. We eat tom[a:touz] and we get a little tense now and then. b. Mozart’s sonatas weren’t for violin and piano, they were for piano and violin. c. I didn’t manage to trap two mongeese: I managed to trap two mongooses.

In (a), the speaker insists on the difference in the pronunciation of tomatoes and the usual way of expressing being under pressure; in (b), the speaker insists that the correct way of describing Mozart’s sonata is the order piano and violin; finally, in (c), what is corrected is an ungrammatical plural form for mongoose. No meaning is at stake, only the correct way to describe things and situations. In contrast, by saying Abi is unmarried, there is no way to mean that a certain linguistic form is inappropriate: indeed, the speaker cannot mean that he cannot say Abi is married. The question is now the following: why is metalinguistic negation restricted to syntactic negation? Why is there no specific linguistic device for marking metalinguistic negative meaning? Why is © unable to do this job? Here are some possible answers. First, as descriptive negation is a pure semantic meaning, reducible to logical negation, and metalinguistic negation a pragmatic use of negation, the difference between two

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 



readings of a negative utterance is mainly a question of contextualization.¹¹ In other words, no linguistic device is required to disambiguate negative utterances. Second, © is not able to have a metalinguistic meaning, because a metarepresentational meaning is a question of usage, and not a question of semantic content. In other words, there is no way to insert, inside a semantic meaning, any device indicating that the expression is not in use but in mention. In the classic distinction between use and mention, the same lexical unit has both usages without any specific formal marking. In effect, there is no specific word, for instance Chicagos, meaning the mention vs. the use of the word Chicago: ()

a. Chicago is a big city. b. Chicago has seven letters.

In other words, what makes a negative predicate specific is that the negative feature cannot have a metalinguistic usage. The third pragmatic conclusion is thus the following: ()

A negative predicate cannot encode a metarepresentational meaning.

.. C

.................................................................................................................................. In this chapter I have proposed a positive definition of what a negative predicate is. Three semantic and three pragmatic properties have been argued for: Semantic properties of negative predicates: a. A negative predicate encodes in its lexical meaning a negative feature; b. only lexical categories having as denotation sets of entities or sets of sets of entities are candidates for negative predicates; c. negative predicates are semantically complete. Pragmatic properties of negative predicates: e. Negative predicates are stronger negative statements than syntactic negative ones; f. negative predicates are used under syntactic negation scope to mitigate the alternative explicitness of the positive predicate, which is conversationally implicated; g. a negative predicate cannot encode a metarepresentational meaning. The conclusion of this chapter is thus the following: negative predicates are motivated both for semantic and pragmatic reasons. The semantic properties explain the nature of the

¹¹ See Horn () for the assumption that natural language does not have a specific metalinguistic operator. See also Blochowiak and Grisot () for an experimental confirmation.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

negative meaning ©, the restriction on © to lexical categories, whose semantics denote sets of entities, and the completeness of the negative predicate semantic meaning (contrary to the meaning of a negative sentence). On the other hand, the pragmatic properties of negative predicates explain why they are stronger than syntactic sentences, the function of negative predicates, and the absence of metarepresentational meaning encoded in negative predicates.

A

.................................................................................................................................. This chapter has been written thanks to the support of the Swiss National Science Foundation for the project LogPrag—Semantics and Pragmatics of Logical Words: Logical Connectives and Negation (project no. _). Many thanks to Joanna Blochowiak for her comments.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

  ......................................................................................................................

 ......................................................................................................................

 

D is something we do; it is a speech act. Negation, on the other hand, is a particular lexical item. Despite being very different kinds of things, denial and negation certainly seem to have something to do with each other. There’s something negative about them both. This ‘negative’ aspect, whatever it is, unifies denial and negation across these categories. It is something that denial does not share with the speech act of assertion, for example, although they are both speech acts; nor does negation share it with, say, ‘must’, although they are both lexical items. There are a range of theories about the relationships between negation and denial. This chapter aims to give a brief overview of these theories, and to indicate some of the reasons for and against each. Before I start, though, some clarifying remarks are in order. ‘Denial’ can mean a lot of things in a lot of contexts, and some of the things theorists have meant by ‘denial’ are not my topic. In particular, the kind of denial I am concerned with is an informative speech act. To deny something is to rule it out, which is to give some information: information about how things aren’t. I am not concerned with senses of ‘denial’ like those invoked in van der Sandt () or Spenader and Maier () that center on removing information from the context of a conversation. And denial is also distinct from what is called ‘weak rejection’ in Incurvati and Schlöder (). This weak rejection is simply an announcement of refraining from commitment. For an example of the kind of thing I am not talking about, consider the ‘no’ in the following dialogue: ()

(a) Alice is in the office. (b) No; she’s either in the office or at home.

The ‘no’ here does not register denial, in my sense, of the claim that Alice is in the office; it simply registers refusal to commit to Alice’s being there.¹ The kind of denial I am interested in is different. It is one which, like assertion, adds information to a conversation. The speech act I here call ‘denial’ is the one called by that name in Restall (), Ripley (b), Price (), Priest (), Dickie (), and Murzi and Hjortland (), and the

¹ See also Stalnaker (: ), who talks of “reject[ing] an assertion” as a way of blocking some of its effects on the context.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

same as the speech act called ‘rejection’ in Incurvati and Schlöder (), Humberstone (), Smiley (), and Price (). (Other authors use ‘rejection’ for a related propositional attitude rather than a speech act, e.g. Restall (), Ripley (b), Priest (), Besson (), and Field (). Rumfitt () seems to have no clear terminological policy, but is about (among other things) the speech act in question.)

.. T  

.................................................................................................................................. Denying a claim A, then, is performing an act that gives some information: information ruling A out. But presumably asserting its negation ¬A is also performing an act that gives information ruling A out. What is the relation, then, between denying some claim A and asserting its negation ¬A? According to at least some theorists, there is no important difference. Call this the equivalence thesis: that denying A is equivalent to asserting ¬A. Depending on what kind of equivalence is being discussed, there are many different versions of the equivalence thesis. For example, someone might think that the important notion of equivalence is having the same effect on the conversational context; someone else might think that the important notion is being coherently performed in exactly the same circumstances; and someone else might think that to deny A is the very same thing as to assert ¬A, that any act that can be correctly described in one way can also be correctly described in the other. These different notions of equivalence are all attempts to get at the pretheoretical sense that there is something importantly the same between denying a sentence and asserting its negation. In any form, the equivalence thesis has come in for some criticism. I’ll return later to some of the reasons for doubting that any version of the equivalence thesis can be made to fly; to begin, however, I’ll take the equivalence thesis for granted. If the equivalence thesis is accepted, there is a question of how it is to be used. Does the equivalence thesis give us a useful theory of what denial is, in terms of negation? Does it give us a useful theory of what negation is, in terms of denial? These are questions of priority or grounding, and they occupy much of the literature on the equivalence thesis.

... As a theory of denial In Frege (/), Frege gives an argument that has long been understood (e.g. in Geach ; Incurvati and Schlöder ) as showing that we should understand denial in terms of negation. In particular, the conclusion standardly attributed to Frege is that denying a claim A is nothing more than asserting its negation ¬A. This is best interpreted as reducing denial to assertion and negation; the latter two are prior, on this view.²

² Rumfitt (: ) doubts this interpretation of Frege, and I do as well. But regardless of what it is that actually happens in Frege (/), something like this argument is standardly attributed to the piece, and is worth exploring here. My point is the argument itself, not whether Frege gave it. In what follows, then, when I say things like “Frege’s argument,” I mean to point to this argument, not to claim that it’s faithful to Frege’s writing.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi





The argument has two steps. One step, based on parsimony and the equivalence thesis, pushes for reduction in one direction or the other. If the equivalence thesis is true, it calls out for explanation. Why should assertion, negation, and denial be related in this particular way? One natural thought is that we should reduce either negation or denial to some combination of the other and assertion. This reduction could then provide an explanation for the equivalence thesis, and it would be a parsimonious explanation as well, reducing away one of the components of the equivalence thesis. (Frege does not consider reducing assertion to denial and negation, although as far as I can see this might also work. See Rumfitt  for related ideas.) If that’s correct, we should be looking to do one of two things. Either we should understand negation in terms of assertion and denial, or else we should understand denial in terms of assertion and negation. But which? The second step of the argument aims to rule out the first option, leaving us to understand denial in terms of assertion and negation. Here’s how it goes. For simple negated sentences, there seems to be not much to tell between the two directions of explanation. Consider the following pair of sentences: ()

a. The accused was in Berlin. b. The accused was not in Berlin.

Here is a simple view of negation on which it can be explained in terms of denial. For ease of reference, I’ll call this the ‘marker-of-denial view’ of negation. On the marker-of-denial view, sincere utterances of (a) and (b) have the very same content, a content according to which the accused was in Berlin. The difference between these sincere utterances is in what they do with this content: a sincere utterance of (a) is an assertion of this content, while a sincere utterance of (b) is a denial of this same content. On this view, the negation in (b) makes no contribution at all to the content expressed in a sincere utterance of this sentence, but rather simply flags the utterance as a denial of that content rather than an assertion of it. This marker-of-denial view contrasts with what is now a more standard view, on which sincere utterances of (a) and (b) can both be assertions. On this view, these assertions involve different contents; negation’s role is to combine with the content expressed by (a) to yield the distinct content expressed by (b), rather than to indicate anything about which speech act is being performed. This is very different from the marker-of-denial view. Frege’s argument shows, to my mind anyhow, that we must accept something like this, and that the marker-of-denial view is untenable. With examples like (a) and (b), it can be hard to see what is at stake between these views; they can even seem like simple redescriptions of each other. Frege’s argument, though, turns on embedding these (in particular (b)) in the antecedent of a conditional, yielding for example (). ()

If the accused was not in Berlin, they are innocent.

In asserting (), a speaker does not either assert or deny that the accused was in Berlin. On the marker-of-denial view, this is hard to explain. If what ‘not’ does is simply to mark denial, why is it not doing its job here? On the other hand, if ‘not’ contributes to the content

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

of the antecedent, then there is no problem; this antecedent is neither asserted nor denied, but rather has a content all its own, a content that contributes to the content of (). So, the argument concludes, in conditional antecedents ‘not’ is not a marker of denial but rather a contributor to content; and without any contravening evidence, we ought to think the same about ‘not’ wherever it occurs. This argument has been considered from a range of perspectives. For example, Smiley (), Rumfitt (), Restall (), Price (), and Humberstone () all discuss it in various ways. I won’t go through these responses here in any detail, but I will note one common thread: commentators basically accept this argument against the marker-of-denial view.³ That is, even theorists who do want to understand denial as prior to (or as giving meaning to) negation do not do so by accepting the marker-of-denial view of negation. In particular, almost all agree with Frege that negation does contribute to content. It is just that this contribution is to be understood in act-theoretic terms, terms that appeal, among other things, to the relation between negation and denial. In going forward, I will assume that this consensus is on the right track: Frege’s argument really does show that the marker-of-denial view is not a correct view of negation. However, as a number of the above-cited works have pointed out, this does not rule out all views on which negation is to be explained in terms of assertion and denial. It simply rules out one such view, the marker-of-denial view.

... As a theory of negation How is it, though, that we can understand negation in terms of assertion and denial without holding to the marker-of-denial view? The answer I’ll sketch appeals to bilateralism: the view that meanings in general are to be given via conditions on assertion and denial. Bilateralism provides an alternative to truth-conditional theories of meaning, and also to assertion-conditional theories of meaning (typically called ‘unilateralist’ theories by bilateralists).⁴ Existing bilateralist theories of meaning can be productively (if roughly) divided into two camps, depending on what kinds of condition on assertion and denial they appeal to. In one camp are bilateralisms like those explored in Price (), Smiley (), and Rumfitt (), which take the relevant conditions to be conditions under which assertions and denials are warranted. These bilateralisms draw closely on unilateralist approaches like those of Dummett () and Prawitz (), which also focus on warrant. On the other hand, there are bilateralisms like those of Restall (), Ripley (), and French (), which take the relevant conditions to be conditions under which whole collections of assertions and denials fit together.⁵ This focus on compatibility fits nicely with

³ However, for an exploration of one way the marker-of-denial view itself might be defended against Frege’s argument, see Bendall (). ⁴ One might imagine a truth-theoretic analogue of bilateralism, appealing to separate truth and falsity conditions; such a thing is suggested for assorted reasons in Barwise and Perry (/), Routley and Routley (), Plumwood (), Priest (), and Kratzer (), although I don’t think the analogy to bilateralism has been explicitly drawn. ⁵ See also in this connection Strawson (: –).

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi





some aspects of Brandom ()’s unilateralist view, and indeed bilateralists of this stripe often draw on Brandom’s work, in a way similar to the way defenders of warrantbased bilateralism drawn on Dummett and Prawitz.⁶ Here, I will refer to those collections of acts that do fit together in the relevant way as ‘in bounds’ and those that don’t as ‘out of bounds’. Warrant-based and bounds-based bilateralisms give a setting for similar-looking theories of the contribution to meaning made by negation: negation swaps assertion and denial conditions. That is, a negation ¬A can be asserted in exactly those circumstances in which the negatum A, the thing negated, can be denied, and denied in exactly those circumstances in which the negatum can be asserted. If meanings are understood in terms of assertion and denial conditions, this is as clear and compositional a theory of negation’s meaning as one could ask for, and it depends directly on denial. Moreover, it immediately yields the equivalence thesis, since it matches the assertion conditions of a negation with the denial conditions of its negatum. The division between the types of bilateralism matters here for what ‘can be asserted’ and ‘can be denied’ mean, and this affects the form of the equivalence thesis that is entailed by this bilateralist theory of negation. For warrant-based bilateralists, the result is that an assertion of ¬A is equivalent to a denial of A in the sense that the two acts are warranted in the same circumstances. For fit-based bilateralists, the result is that the two acts fit together with the same collections of acts. Either way, some form of the equivalence thesis is entailed by the view of negation’s assertion and denial conditions, but the particular form depends on what these conditions themselves amount to. However the equivalence is understood, bilateralisms of this sort give a view on which negation contributes to the content of the clauses in which it occurs. These views are thus not the marker-of-denial view. This is enough to be compatible with Frege’s argument. Despite this, they provide a clear sense in which negation is best understood by appeal to (among other things) denial. On this kind of bilateralist view, semantic content in general is understood in terms of its relations to assertion and denial conditions. Negation’s connection to denial is just a special case of this general principle.⁷ (It is a particularly simple special case, given the ‘swap’ theory of negation’s content entertained here.)

⁶ I discuss these two bilateralist camps and the differences between them in more detail in Ripley (). There are also views very similar to bilateralisms in their logical and ideological underpinnings that don’t work directly in terms of assertion and denial, but rather in terms of conclusion and premise, or in terms of proof and refutation. Examples include Schroeder-Heister (); Tennant (); Wansing (). ⁷ This answers, or at least suggests an answer to, one challenge to bilateralism posed in Humberstone (: ff.). After outlining a hypothetical speech act of “alterjection” that obeys its own form of the equivalence thesis wrt disjunction—so that asserting the disjunction of A with B is equivalent to alterjecting with respect to A and B—Humberstone challenges the bilateralist to “show how the claim for the conceptual priority of rejection over negation is any more plausible than the corresponding claim for the conceptual priority of alterjection over disjunction” (). The bilateralist’s response should be: that’s not the appropriate correspondence. The sense in which denial is prior to negation corresponds to the sense in which denial (not alterjection) is prior to disjunction as well. (For more on the relations between denial and disjunction in a bilateralist setting, see Ripley ().)

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

.... The other equivalence thesis It’s worth noting another commitment that this ‘swap’ theory of negation’s meaning takes on: that denying a negation is equivalent (in whatever sense is appropriate) to asserting its negatum. This claim—call it the ‘other equivalence thesis’—is certainly reminiscent of the equivalence thesis itself, but it is independent. Indeed, nothing at all follows directly from the equivalence thesis about how the denial of a negation relates to acts involving its negatum. But as bilateralists understand meaning in terms of assertion and denial conditions, they are obligated to say something about the conditions under which negations may be denied, at least if they are attempting to give a full theory of negation’s meaning. (The swap theory, which is the only option I’ll consider here, is the usual option, but there is certainly room within a bilateralist framework for other theories of negation’s meaning, including theories that don’t entail the other equivalence thesis. For debate about the other equivalence thesis in this kind of setting, see Price ; Price n.d.; Rumfitt .) There is a way, however, to appeal to the equivalence thesis, together with at-leastsomewhat plausible background assumptions, in defense of the other equivalence thesis. This has been explored directly in a fit-based bilateralist setting, so I will reproduce the reasoning in that setting. (The reasoning here is essentially that of Sambin, Battilotti, and Faggian (), as deployed in Restall n.d.) Suppose three things: first, that any collection containing an assertion and a denial of the very same content is out of bounds; second, that if a collection of acts is out of bounds, then any collection containing that original collection is also out of bounds; and third, that for any collection C of acts that is in bounds and any content A, at least one of the following two collections of acts must be in bounds: either the collection that results from adding an assertion of A to C, or the collection that results from adding a denial of A to C. The first two of these assumptions are less complicated and easier to justify: the first amounts simply to the claim that assertions and denials of the same content clash with each other directly; and the second to the claim that one can’t undo a clash by making more assertions and denials.⁸ The third assumption is more complex, but there are at least worldviews from which it seems plausible. It tells us that there are no ‘double binds’: no in-bounds collections of acts that simultaneously rule out assertion of A and rule out denial of A, for any content A.⁹ Now suppose the equivalence thesis, in the form appropriate for bounds-based bilateralists: that an assertion of ¬A fits with the same collections of acts as a denial of A does. From all this, we can demonstrate the other equivalence thesis. In the appropriate form, we are aiming to show that a denial of ¬A fits with the same collections of acts as an assertion of A does. To see this, first suppose we have some collection C of acts with which a denial of ¬A fits. I’ll argue that an assertion of A must also fit with C. Note that, by our first assumption, C together with both a denial and an assertion of ¬A is out of bounds. So, by the ⁸ This second claim, then, records the idea that we are dealing with informative denials, not simply retractions. ⁹ See Restall (, ) and Ripley (, a) for discussion of this third assumption. The third assumption can be strengthened in ways that allow us to dispense with the second for the purposes of the following argument, but at the cost of additional complexity; it would be a distraction here, although it would bring us closer to Sambin, Battilotti, and Faggian ()’s formulation of the argument.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi





equivalence thesis, C together with a denial of ¬A and a denial of A is out of bounds. By our third assumption, though, we must be able to add either an assertion or a denial of A to C together with a denial of ¬A. Since we can’t add a denial of A, as we’ve seen, we must be able to add an assertion of A. So C together with a denial of ¬A and an assertion of A is in bounds. But then by our second assumption, C together with an assertion of A is in bounds too. To see the other direction, suppose the reverse: that we have some collection of acts D with which an assertion of A fits. I’ll argue that a denial of ¬A must also fit with D. Note that, by our first assumption, D together with both an assertion and a denial of A is out of bounds. So, by the equivalence thesis, D together with an assertion of A and an assertion of ¬A is out of bounds. By our third assumption, though, we must be able to add either an assertion or a denial of ¬A to D together with an assertion of A. Since we can’t add an assertion of ¬A, as we’ve seen, we must be able to add a denial of ¬A. So D together with an assertion of A and a denial of ¬A is in bounds. But then by our second assumption, D together with a denial of ¬A is in bounds too. None of the three assumptions drawn on in this argument is uncontentious, but neither is any of them obviously wrong. If they do hold, then the equivalence thesis and the other equivalence thesis cannot come apart; otherwise, they might.¹⁰

.... Classifying token acts Bilateralist approaches, at least those that hold to the ‘swap’ theory of negation’s content, show how to maintain that denial is prior to negation in a way that Frege’s argument doesn’t refute. The bilateralist holds that negation is a genuine operation on the content of a sentence; but this is all Frege’s argument really establishes. So the bilateralist joins with Frege in rejecting the marker-of-denial view of negation, but maintains that denial is nonetheless prior to negation.¹¹ It’s just not prior in the particular way Frege argues against. A natural question arises here, though: what are we to say about sincere utterances of (b) (‘The accused was not in Berlin’) and the like? Are these all assertions that the accused was not in Berlin? All denials that the accused was in Berlin? It looks like the bilateralist shouldn’t maintain either of these uniform views. After all, if all such utterances are assertions of the negated content, then denial as a separate speech act threatens to disappear entirely—but the bilateralist needs denial to play a key theoretical role, so it would be a problem if it’s never actually observed. On the other hand, if these are all denials of the unnegated content, it’s not clear why this would be so. Given that there is a negated content (as Frege’s argument shows, and as the bilateralist agrees), what obstacle could there be to asserting it? There are two plausible ways for a bilateralist to respond to this challenge, I think. The first is to maintain that some of these utterances are assertions and others denials. This would avoid both problems floated above that the uniform views face: since some of these ¹⁰ By swapping “assertion” and “denial” in the above arguments, you can show from the same assumptions that the other equivalence thesis implies the equivalence thesis as well. ¹¹ As Price () puts the point, Frege’s argument “is not an objection to treating denial as something other than the assertion of a negation; but rather to treating the negation sign as nothing other than an indication that a sentence in which it occurs has the force of a denial” ().

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

utterances are denials, denials don’t disappear from view; and since some are assertions, there is no absence of assertions to explain. Bilateralists pursuing this option would need to provide some story about what the difference is between those utterances that are assertions and those that are denials, but I don’t think there are strong reasons to suppose this can’t be done. The second option, and the one I find more plausible, is to hold to both uniform views. Every sincere utterance of (b) is at the same time an assertion that the accused was not in Berlin and a denial that they were. Price () suggests this option, pointing out that to take it up would be to reject the idea “that the meaning of every utterance should have a unique resolution into a component due to sense, and a component due to force” (). This view too avoids the troubles that threatened each uniform view on its own: there are plenty of denials and assertions around to answer both worries, if each sincere utterance of a negated sentence is simultaneously one of each. This option raises a further question: if sincere utterances of (b) are simultaneously both assertions that the accused was not in Berlin and denials that they were, then what about sincere utterances of (a) (‘The accused was in Berlin’)? Are these too simultaneously both assertions that the accused was in Berlin and denials that they weren’t? I think so. Every speech act that is either an assertion or denial is in fact both. (This requires both the equivalence thesis and the other equivalence thesis, or at least fits most cleanly with holding to both.) Assertions and denials seek to inform by telling us at the same time how things are and how things are not, by ruling in and ruling out simultaneously. Forms of this idea, I think, can also be found in Frege (/: ) (“When it is a question of whether some thought is true, we are poised between opposite thoughts, and the same act which recognizes one of them as true recognizes the other as false”) and Strawson (: ) (“For when we say what a thing is like, we not only compare it with other things, we also distinguish it from other things. (These are not two activities, but two aspects of the same activity)”). See also Rumfitt () for more examples from Frege on this score. (This claim—that every assertion of a negation is a denial of the negatum—is what’s called the “Equivalence Thesis” in Parsons . This is a particularly strong form of what I’ve been calling the “equivalence thesis,” since identity is a particularly strong form of equivalence. Parsons also notes that the other equivalence thesis is an optional add-on.)

.. A   

.................................................................................................................................. So much for views that hold to the equivalence thesis. On the other side of the aisle, there are views about negation and denial on which asserting ¬A and denying A can be quite different indeed. Often, this comes packaged with the idea that some claims are neither true nor false. The idea is that since these claims are not true they should be denied, to indicate this; but since they are not false their negations should not be asserted. A classic statement of this idea is Parsons () (using “reject” instead of “deny”): [S]ometimes when we “say something negative” we should not be thereby committed to an assertion of a negative claim, for we are not asserting at all, we are only rejecting something. I might say “Paul Bunyan is not bald” without thereby committing myself to the truth of the sentence ‘Paul Bunyan is not bald’, for I might think (as many people do think) that this

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi





sentence lacks truth-value. . . . Sometimes saying a sentence with a negative word in a certain tone of voice just counts as a rejection of the corresponding positive version. ()

Parsons also considers in a similar light, “The purpose of life is not to serve mankind; it doesn’t make sense to ascribe purpose to life.” There is a link here to “metalinguistic negation” (for which see Horn ; Geurts ; Carston ; and Chapter  in this volume), and relatedly to long-standing arguments about whether negation is ambiguous (for which see Horn ; Atlas ; Marques ; Kempson ). Indeed, Tappenden () seems to take a wide range of standard cases of metalinguistic negation to provide counterexamples to the equivalence thesis. He describes these as cases in which “the use of ‘not’ indicates the rejection of a candidate assertion, but clearly not the assertion of the negation of the sentence in question” (). As examples, Tappenden gives all of: ()

a. Old Liz did not wave at you. Queen Elizabeth the Second waved at you. b. John isn’t wily or crazy. He’s wily and crazy. c. Ruth didn’t manage to solve the problem. She solved it with ease.

Here, (a) is naturally understood as rejecting the choice of ‘Old Liz’ to refer to Queen Elizabeth; (b) as rejecting the implicature carried by ‘John is wily or crazy’ that John isn’t both; and (c) as rejecting the presupposition ‘manage’ might carry that the problem was difficult for Ruth to solve.

... Frege revisited As Parsons and Tappenden present them, these examples seem to be intended as obvious counterexamples to the equivalence thesis. But that they are counterexamples is not obvious at all. In this section, I’ll present an argument for the conclusion that these instances of metalinguistic negation are not best understood as simply marking denial of the corresponding positive content. The argument that shows this is just Frege’s again. Parsons’s and Tappenden’s view of these uses of negation is precisely the marker-of-denial view, but we’ve already seen that Frege’s argument gives us a powerful reason to reject the marker-of-denial view. Although Parsons and Tappenden do not hold the full marker-of-denial view, they do hold the view as applied to these cases. So as long as Frege’s argument can be specialized to these cases, it works just as well against Parsons’s and Tappenden’s views as it does against the markerof-denial view full stop. And indeed, these metalinguistic uses of negation work just fine under embeddings, where they are not plausibly understood as encoding any kind of speech act. Consider, for example: ()

a. If that’s not Old Liz waving at us, then that’s presumably not Prince Dickhead standing next to her. b. If John isn’t wily or crazy, but wily and crazy, then we may have a problem on our hands. c. If Ruth didn’t manage to solve the problem, but solved it with ease, then I’ve underestimated her.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

Someone who asserts (b), for example, hasn’t denied that John is wily or crazy. But whatever the negation in ‘John isn’t wily or crazy’ is doing in (b), it’s the same thing it’s doing in (b). So the negation in (b) is not an indicator of denial either. The same argument works for the other examples as well. If this argument is on the right track, we should not understand the distinctive behavior of metalinguistic negations like these in terms of their indicating a speech act of denial, since they retain their distinctive behavior even in the antecedents of conditionals, where it doesn’t seem that they can possibly be indicating such a speech act.¹² For related discussion, see Carston (), Carston and Noh (), Chapman (), Geurts (), Giannakidou and Yoon (), Kempson (: ), Larrivée (), and Textor (). On the other hand, if the views put forward by Parsons and Tappenden can be successfully defended from this argument, it would be worth revisiting more global applications of Frege’s argument in light of that defense. If the marker-of-denial view can be made to work in these limited cases, even though the negations involved embed without difficulty into the antecedents of conditionals, then perhaps it can be made to work for negation across the board. I close this section by mentioning that both Parsons and Tappenden reject the equivalence thesis in the service of understanding paradoxical sentences, in particular taking account of the possibility that such sentences might either be such that neither they nor their negations are true, or be such that both they and their negations are true. On the first possibility, their idea is that both would be deniable but neither assertible, and on the second, their idea is that both would be assertible but neither deniable. Either way, both the equivalence thesis and the other equivalence thesis would be violated. So there remains a possibility of using the paradoxical sentences themselves in an argument against the equivalence thesis. Indeed, this is what Priest (: ch. ) does. (Priest mentions some metalinguistic-negation-involving cases, but does not rest his case on them, preferring to focus directly on the paradoxes.) Essentially the same argument, however, can be used here; the question of embedding the alleged ‘marker’ uses of negation arises just the same in paradoxical cases as in other cases. See Shapiro () and Ripley (b) for further discussion of this issue as it applies to paradoxical sentences in particular.

... Suspicious similarities In this final section, I present an argument sketched in Ripley (b) for the equivalence thesis. It turns on substantive assumptions about the interactions between negation, conjunction, and disjunction, and between denial, conjunction, and disjunction, so is hardly theoretically neutral. But the assumptions it draws on are shared by those who reject the equivalence thesis, at least those I’ve discussed here, so I think it remains a fair argument.

¹² Tappenden () comes very close to addressing this objection, even mentioning “If Ruth didn’t manage to solve the problem, she must have solved it with ease” (). But the response offered there is simply “it is not maintained here that the sole function of ‘not’ is to indicate a speech act” (, emphasis in original). The argument I have just presented, however, does not turn on attributing any such claim to Tappenden.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi





The argument starts by noting a range of similarities between these interactions. For example, someone who denies a disjunction may as well have denied one disjunct and then the other. And someone who asserts the negation of a disjunction may as well have asserted the negation of one disjunct and then the negation of the other. For another example, someone who denies a conjunct of a conjunction seems committed to a denial of the conjunction as well. And someone who asserts the negation of a conjunct of a conjunction seems committed to an assertion of the negation of the conjunction as well. What explains these similarities between denial, on the one hand, and assertion of negation, on the other? If the equivalence thesis holds, we have a very direct explanation of the similarities: they follow straightforwardly from the equivalence thesis. But if the equivalence thesis fails, that explanation doesn’t work. We would need some other explanation. Perhaps it is simply a coincidence, and no explanation at all is needed. But when there is an explanation as simple as the equivalence thesis available, it is difficult to be happy with that result. As far as I know, however, nobody who rejects the equivalence thesis has undertaken to provide an explanation of the sort that would help here. So there would seem to be an explanatory gap, at least for now, in those theories that reject the equivalence thesis.

.. C

.................................................................................................................................. Discussion of the relationships between negation and denial has tended to center on the equivalence thesis: whether it is true, and if it is true which way the direction of priority runs. While Frege’s argument is sometimes claimed to show that negation must be prior to denial, in fact its conclusion is narrower, merely ruling out the marker-of-denial view of negation. This argument does not settle the priority question in either direction. Bilateralist views of content can accept the equivalence thesis, and hold that denial is prior to negation, without accepting the marker-of-denial view of negation; Frege’s argument thus does not refute them. On the other hand, at least some arguments that have been offered against the equivalence thesis do seem to run into trouble from Frege’s argument, since they involve the marker-of-denial view of at least some uses of negation, even if not all. Moreover, the equivalence thesis provides a simple explanation for some otherwise puzzling parallels between negation and denial.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  ......................................................................................................................

             ......................................................................................................................

  

.. I

.................................................................................................................................. K’ () well-known distinction between sentence and constituent negation has determined the syntactic and semantic literature on negation over recent decades. The distinction is based on four well-known syntactic tests: the either/too-test, (), the not eventest, (), the question tag test, (), and the neither-test, (). Sentences that can combine with (n)either, positive question tags, and not even give rise to sentence negation, cf. the a-examples, whereas those that cannot are either affirmative, cf. the b- examples, or consist of a constituent negator, which is a broad term encompassing also cases of affixal negation, cf. the c-examples (Klima : –/). ()

a. Publishers will usually/always/not reject suggestions, and writers will not/scarcely/ hardly/never/seldom/rarely accept them either/*too. b. Writers will never accept suggestions, and publishers will always/surely/usually/ commonly reject them *either/too. c. *Publishers will unintentionally reject suggestions, and writers will unintentionally reject them either.

()

a. The writer will not/never/seldom/rarely accept suggestions, not even reasonable ones. b. *The publisher often/commonly disregards suggestions, not even reasonable ones. c. *Writers are unreceptive to suggestions, not even good ones.

()

a. Writers will never accept suggestions, will they/ *won’t they? b. Publishers will surely reject suggestions, *will they/ won’t they/ will they not? c. *Writers unfortunately reject suggestions, do they?

()

a. Writers won’t be accepting suggestions, and neither will publishers. b. *Writers will always be accepting suggestions, and neither will publishers. c. *Writers will unfortunately always be accepting suggestions, and neither will publishers.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



Basically, the Klima-tests distinguish between negative markers that have wide scope over the tensed predicate and negative markers that do not have scope over the tensed predicate and hence can be said to have low(er) negative scope. A similar distinction as the one found in Klima can be found in the work by Jespersen (). He referred to sentence negation as nexal negation and to all other types of negation as special negation. The term nexal refers to the fact that a negative form unites two different “ideas” (Jespersen : ), as in (). In this sentence the “idea” he and the idea coming are “negatived” by the nexus n’t. ()

He doesn’t come.

Jespersen also points to the fact that the distinction between special and nexal negation is clear in principle but that there are instances which are ambiguous. He discusses the example in (a), which displays nexal negation, but which receives a special negation reading due to the addition of the only-phrase in (b). ()

a. He doesn’t smoke cigars. b. He doesn’t smoke cigars, only cigarettes.

This distinction between sentence and constituent negation or nexal and special negation can actually be traced back even further. Aristotle namely also distinguished between term negation, (a) and predicate denial, (b). ()

a. It is a not-white log b. It is not a white log. (From Aristotle, Prior Analytics, cited in Horn a: )

But how do these differences relate to the other well-known dichotomy, that is the dichotomy between internal and external negation? The distinction between internal and external negation goes back to the Stoics, who were the first to develop a propositional logic (Horn a: ). Only negation that was literally outside of or external to the clause was considered propositional negation, (a), whereas negation as in (b) was considered internal to the proposition. ()

a. not: it is day and light b. It is day and it is not light.

(Horn a: )

It was the Stoic school that influenced Fregean propositional logic and that has had an enormous influence on the development of formal semantics and also on the treatment of sentence negation as a propositional negator since then. What is clear is that the external negation in (a) differs from Aristotle’s predicate denial or Klima’s sentence negation, which are both more like the internal negation in (b). Actually, Horn (a: ) argues—relying on typological work by Dahl ()—that there is no empirical evidence for an external negator (like the one in (a)) in natural language, i.e. the idea of an external negator is a construct by the advocates of propositional logic. Horn therefore argues for a return to Aristotle’s predicate and term logic. However, in recent work Bar-Asher Siegal (a, b) argued for the existence of an external negator in Jewish Babylonian Aramaic (JBA), the negator lāw. If the

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



  

Table .. Terminological overview  n’t/not external negation Klima Frege Aristotle

non-

un-/iN-/ dis-, a-

internal negation sentence negation

constituent negation

propositional negation predicate denial

Horn/Dahl Jespersen

not

predicate term negation morphological/lexical negation

nexal negation

special negation

identification of this negator as an external negator is correct, then this runs against the established idea that there is no such thing as an external negator in natural language. More discussion of this negator is taken up in section .. Within the frame of the Stoics, Aristotle’s, Jespersen’s and Klima’s distinction between the two scopal types of negative markers can only be understood as two different types of internal negations. Notwithstanding the terminological disagreement, we end up with more types of negation than either dichotomy mentions, that is an external negator and at least three types of internal negation: Klima’s sentence negation, Klima’s constituent negation, and the privative or affixal negation that this latter group also includes, which is treated as an independent type of negation by Aristotle and the Stoics. Other terms for affixal negation are “morphological negation” (Horn a: ; Hamawand ) or “lexical negation” (Dahl ). Table . summarizes the terminological distinctions discussed up until now with examples from English in the top row. The slot for external negation is left blank, though it may very well be that this could be filled by JBA lāw for instance (cf. section .). Summarizing our discussion so far, different terminology has been used to refer to wide and narrow scope negators. Two dominant dichotomies have governed the discussions on negation for decades: sentence negation vs. constituent negation and external negation vs. internal negation. Our introduction showed that on the one hand there is overlap between these terms, and on the other hand there are problems related to these terms, that is one term (external negation) has been questioned for its empirical significance and the other terms have been used to refer to various different types of negation. In section . we will consider in more detail some of the tests which led to the two dominant dichotomies, hence also further clarifying the meaning of these different terms. We will discuss some caveats and problems with the tests, pointing to the most problematic issue, that is that none of the tests is capable of distinguishing different types of constituent negation or low scope negation. This problem is actually already apparent in the terminological overview in Table .: the two main dichotomies do not deal with subdivisions within the group of low scope markers, that is with the differences between the English low scope not, non-, and un- for instance. In section . we will take a closer look at De Clercq (, b) and the findings of De Clercq (, to appear), who studied the morphology of negation and the stacking properties of negative markers to distinguish the different types of negative markers a language has at its disposal, including differences between low

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



scope negative markers. Before we proceed, I want to point out that the focus of this chapter will be on negative particles modifying a predicate, predicate term, or proposition, to a lesser extent on negative quantifiers (cf. Chapter  in this volume) or scalar quantifiers, and not on Negative Polarity Items (NPIs) (see Chapter  in this volume).

.. T  : A  

.................................................................................................................................. As we discussed in the introduction, the dichotomy between sentence and constituent negation is closely associated with Klima’s work. Naturally, the Klima tests were challenged and critically reviewed by many researchers, and other tests for sentence negation were proposed. In what follows we will review some of these tests and critical discussions. Our investigation will lead to the conclusion that the dichotomy made by Klima on the basis of his tests is a valid and useful dichotomy. However, we will argue that some caveats need to be kept in mind when applying the tests. In the course of the discussion, it will become clear that some criticism on the tests is actually related to terminological confusion having to do with the other dichotomy we started out with, that is external vs. internal negation, and with the original aim of Klima’s tests. One of the first syntacticians to challenge the Klima-tests was Jackendoff (: ). He argues that real sentence negation can be defined as in (): ()

A sentence [S X-neg-Y] is an instance of sentence negation if there exists a paraphrase It is not so that [S X-Y].

According to Jackendoff, not all sentences that pass the Klima-tests for sentence negation give rise to what he defines as sentence negation in (). A problematic pair is in (). () a. Many of the arrows didn’t hit the target. b. Not many of the arrows hit the target.

(Jackendoff : )

The data in () illustrate that the sentences in () both pass the Klima neither-tests for sentence negation (cf. De Haan : – for more discussion). ()

a. Many of the arrows didn’t hit the target, and neither did many of the javelins.1 b. Not many of the arrows hit the target, and neither did many of the titanium-alloy bullets. (Jackendoff : )

However, (b) can be paraphrased with the ‘Jackendoff test’ (see de Haan  for that label), that is with It is not so that X-Y, whilst this is not the case for the sentence with the

¹ Jackendoff () presents (a) with two question marks, but De Haan (: , fn. ) argues that several of his informants considered the neither-tag acceptable.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



  

regular negation of the auxiliary in (a). Due to the fact that the quantified subject is not negated in (a), Jackendoff argues that Aux-negation does not give rise to real sentence negation but only to the negation of the VP. He adduces the test in () in support of that claim. ()

a. # Not many of the arrows hit the target, but many of them did hit it. b. Many of the arrows didn’t hit the target, but many of them did hit it.

In other words, the scope of negation is not the full sentence in (a), whilst it is in (b). For Jackendoff it thus seems that only a sentence in which negation takes scope over the subject is real sentence negation. Aux-negation is under his proposal not necessarily real sentence negation. He refers to it as VP-negation, (a), and considers it some sort of constituent negation. De Haan () refers to the Klima-tests as syntactic tests for sentence negation, whilst he refers to Jackendoff ’s test as a semantic test for sentence negation. De Haan points out that “The Jackendoff test is [ . . . ] sensitive to the relative scope of negation and quantifiers,” but that it is not a test for sentence negation. If we were to follow Jackendoff ’s proposal for what sentence negation is, that is a negative sentence that can be paraphrased by It is not the case that, then the pair in (a)–(b) could both be considered instances of sentence negation, because both (a) and (b) can be paraphrased as (c). However, it is quite obvious that the un- in unhappy does not have the same scope according to the Klima-tests. ()

a. That actor is not a professional. b. That actor is nonprofessional. c. It is not so that he is (a) professional.

De Haan (: ) concludes that the Klima-tests and the Jackendoff-test do different things: the difference is a difference between semantic and syntactic testing of sentence negation. I agree with this observation, and hence point to the fact that the term ‘sentence negation’ is also an umbrella term for different types of negation. The Jackendoff-test is indeed capable of detecting which negation can give rise to external negation, that is to scope over the entire sentence, including the subject. However, the test massively overgeneralizes, as illustrated by the examples in (). The Klima-tests yield a positive result for sentence negation if a specific negative marker is involved (n’t or not in English), which takes scope in a sufficiently high position. However, it seems that several syntactic positions are involved in what is referred to as sentence negation. Whilst sentence negation usually gives rise to external negation, it can fail to do so in the presence of another operator, as was the case in (a), strongly suggesting that the operator many intervenes between two positions for sentence negation. We get back to this below. De Haan (: ) concludes furthermore that all Klima-tests are reliable tests for sentence negation, which even seem to have cross-linguistic validity. Kraak () and Seuren () studied the tests for Dutch, Stickel () for German, Attal () for French, and Ibañez () for Spanish. De Haan himself also looks at a couple of non-Indo-European languages (Yoruba, Yavapi, etc.) to provide support for the crosslinguistic validity of the Klima tests for sentence negation. Only the question tag-test, he

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



argues, is an English-specific test, for which the cross-linguistic validity has not yet been established. McCawley (b: ) pointed to the question tag-test as a potentially problematic test due to the fact that there are two types of tags: question tags and reduplicative tags. Reduplicative tags reduplicate the polarity of the clause they are appended to. Positive reduplicative tags combine with positive sentences, (a). Negative reduplicative tags are said to be non-existent, (b). ()

a. John bought the book, did/*didn’t he? (Aff-S) b. John didn’t buy the book, *did/*didn’t he?

Without any context, a positive tag could thus also be a sign of a positive sentence, leading to confusion according to McCawley (b: ). However, apart from a distinction by means of intonation patterns (Ladd ; Quirk et al. ), question tags and reduplicative tags can be kept apart thanks to the Oh so-test (Quirk et al. : –; De Clercq ). Sentences with reduplicative tags can be preceded by Oh so; as such the speaker’s suspicion is signaled (see also Chapter  for a discussion of how different types of question tags can be pragmatically modeled), whilst this is not possible for question tags, which merely check for information. Moreover, since negative tags cannot be reduplicative tags, De Clercq () points out that they can safely be considered as an indication of the positive nature of the main clause. Only with positive tags should some semantic testing be done to assure oneself of the type of tag that is being used, that is a reduplicative or question tag. Picking up on McCawley’s (b) worry, Brasoveanu et al. () also argue for the validity of the question tag-test on the basis of an experiment. In addition, their experiment also sheds some light on the nature of negative subjects, a topic that we have already briefly introduced in relation to (b). Finally, their experiment also clarifies some aspects of the interaction between the question tag-test and scalar quantifiers, an issue that we will come back to below. With respect to their first point, namely the validity of the question tag-test, we can say that the results of their experiment suggested that—provided the right context is given to informants—question tags can be triggered instead of reduplicative tags. The context created to trigger question tags was a context that matches well with the idea of information checking. Each test sentence was preceded by a scene-setting sentence followed by “Aha I was right!,” illustrated for a positive and a negative sentence in () and () respectively (Brasoveanu et al. : ). ()

Mary has just told Jane something about government representatives. Jane knows it already and says: “Aha, I was right! The government representative visited the colonies this year, did he/didn’t he?”

()

Geoff has just told Ryan something about composers. Ryan knows it already and says: “Aha, I was right! The composer did not use the cello in his late period, did he/didn’t he?”

In the experiment, the negativity of items like no, never, few, and rarely was checked. As controls, affirmative sentences and sentences with not were used. The results indicated that the question tag test worked well: the items which triggered the highest number of positive

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



  

tags (i.e. indication of negativity) were never, not, and no, followed by few and rarely. With respect to the second point addressed in Brasoveanu et al. (), the nature of negative subjects, the chapter also confirms that negative subjects tend to trigger sentence negation more often than negative objects. This fact has been referred to as the subject-object asymmetry in the literature (Beghelli ; De Clercq ; De Clercq, Haegeman, and Lohndal ). Sentences with negative subjects (no X) predominantly give rise to positive question tags, whilst sentences with negative objects do so much less. Summarizing, the reservations uttered with respect to the question tag test can be discarded thanks to this experiment. In addition, the experiment confirms that negative subjects are “more” negative than negative objects, a result that matches well with the idea introduced in relation to (b) that negative subjects give rise to external negation. In Penka’s () discussion of the Klima-tests she considers it a problem that Klima’s tests are not only applicable to negative indefinites and not, but that they also single out few, seldom, rarely, etc. as yielding sentence negation. The latter elements are only “downward entailing” (Ladusaw ; henceforth DE), i.e. they support inferences from set to subset, and are not negative according to Penka (: ), that is, not “antiveridical” (Giannakidou ), “antimorphic” or “antiadditive” (van der Wouden : ). Penka’s idea squares with Beghelli’s () in this respect, who refers to the latter group of elements as non-negative downward entailing elements. As such, Penka does not consider the Klima-tests a foolproof test for real sentence negation. However, even though I agree that this fact adds a (semantic) caveat to the tests, it does not invalidate Klima’s original aim and claim: the tests’ effectiveness in distinguishing between the constituent scope or sentential scope of an abstract [neg]. To the contrary, Brasoveanu et al. (), whose experiment we introduced above, also showed that the semantic hierarchy of negativity observed in the literature on negation and NPI-licensing is maintained after application of Klima’s question tag test, that is downward entailing operators like few and rarely do not yield positive question tags (i.e. Neg-S) as often as for instance no, never, or not. However, the fact that they are sometimes compatible with positive question tags could show that an abstract element [neg] must be present underlyingly, which was exactly what Klima’s tests were designed to identify. Support for this idea comes from De Clercq (), who shows how DE expressions like squatitive negation cannot trigger positive question tags and hence cannot give rise to sentence negation in Klima’s terms, (). ()

a. Janet read squat, *did she/didn’t she? (Aff-S) b. Janet read fuck-all books, *did she/didn’t she? (Aff-S)

(De Clercq : )

These squatitive negative elements are also not capable of licensing NPIs, a fact observed by Postal (a: ). It thus seems that there are downward entailing expressions that do not contain an abstract [neg], and that can only give rise to semantic (sentential) negation (cf. Zeijlstra b), such as squatitive elements. However, few and rarely, etc., contain [neg] and are scalar and downward entailing. More support for the presence of [neg] in few comes from typological work on few, that shows that in quite some languages few can only be expressed by means of an overt negation and another element, hence indirectly providing support for a decomposition of few into [neg] and an another element (De Clercq a). Another argument in support of the Klima tests comes from interactions with negatively prefixed expressions. These expressions are downward entailing and can license NPIs since

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



they consist of a [neg] feature, (), but the Klima-tests correctly classify them as giving rise to Aff-S, (). ()

()

a. b. c. d.

He is unable to find any time for that. It is inconceivable that he could do any more. It is impossible/unlikely that he will do any more. He would be unwise to do any more.

(Klima : –)

a. He is unable to find any time for that, isn’t he? b. He would be unwise to do any more, wouldn’t he?

Data like these show that the Klima-tests are capable of capturing scope distinctions between markers endowed with [neg], but that they do not distinguish between low scope negativity and affirmation. Another issue with the Klima-tests was identified by Payne (: ) and further discussed by Penka (: ). Payne adduced data in which an adverbial quantifier intervenes between the subject and the negated auxiliary, (b), yielding an affirmative sentence instead of a negative sentence, (a). ()

a. John doesn’t often pay taxes, does he/*doesn’t he? b. John often doesn’t pay taxes, *does he/doesn’t he?

On the basis of this, Penka argues that it is the widest scope-bearing element which the question tags are sensitive to, and not so much the negation itself. De Clercq () follows this line of thought, claiming that it is ‘sentential polarity’ that the tags (and other Klimatests) are sensitive to, and not so much sentential negation. De Clercq () provides a syntactic account, in which she argues that the regular position for sentential negation is a position at the level of the tensed predicate, that is, at TP in syntactic terms (see also section .), whilst sentential polarity resides in a left peripheral position in the clause (possibly a focus position, see section .), from where it determines the polarity of the clause (De Clercq ; De Clercq, Haegeman, and Lohndal ). The idea is that this position needs to be checked or filled, covertly or overtly, to type the polarity of the clause. If negation outscopes the tensed verb and in the absence of another scope-bearing element, the Klima-tests in general and the question tags in particular will yield a sententially negative sentence, that is a sentence with sentential negative polarity. The interpretation of this negation will be an external negation. However, if a scope-bearing positive element is present in between the position dedicated to sentential polarity and the position for sentential negation at the tensed predicate, this element will intervene between the overt negation and the sentential polarity position, so that the polarity of the clause will be affirmative. An external negation reading will not be possible then. The idea that sentential negative polarity resides somewhere (high up) in the left periphery is also present in the work of Moscati (, ) and McCloskey (), who take the perspective that negative clause typing happens at the level of the complementizer, that is at CP, for which languages with negative complementizers provide extra support. Penka’s () definition of sentence negation actually incorporates the idea that there are different levels of being sententially negative. This idea is present in the use of the word at least in her definition of sentence negation in ().

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



  

()

Negation taking scope at least above (the existential quantifier binding the event argument of) the main predicate.

Her approach is in line with work by Davidson () and Acquaviva (), who argue that sentential negation arises when negation takes scope above the event expressed by the verb. For a sentence as in (a) the negation outscopes the existential quantifier, which binds the event of the main predicate, cf. (b) (Penka : ). ()

a. John didn’t kiss Mary. b. ¬ 9 e [agent(John, e) & theme(Mary, e) & kiss(e)]

Even when there is a quantifier intervening between negation and the existential quantifier, sentential negation can still arise, in spite of the fact that there will not be an immediate scopal relation between negation and the existential quantifier, as illustrated for (a) in (b). ()

a. Not every boy kissed Mary. b. ¬∀ x [boy (x) → 9e [agent(x,e) & theme(Mary,e) & kiss (e)]]

() is reminiscent of the example in (b), and hence also illustrates an external negation reading. However, given the definition provided in (), also sentences as in () should be considered instances of sentence negation. Also in () the scope of the negation is higher than the existential quantifier that binds the event expressed by the main predicate. However, the scope of the negation in () is not external nor does it affect the tensed auxiliary. ()

She has not run to the shop, but walked there.

Penka addresses this issue briefly but decides to remain agnostic with respect to the question whether sentence negation takes scope over tense or below tense. The idea that sentence negation is a concept referring to multiple types of negation and that it arises whenever negation takes scope in one of the many positions above VP is also present in syntactic work by Zanuttini (). Zanuttini argues on the basis of dialectal research in Italian that there are at least four different syntactic positions within the clause where sentence negation can appear. Poletto (, ) follows up on this and argues that these different positions are associated with etymologically different types of negative markers (see Chapter  in this volume). However, it is unclear how these different positions relate to the two dichotomies under discussion in this chapter. In our survey of the discussions related to the two dichotomies we started out with, sentence negation is most discussed. The upshot of the entire discussion is that sentence negation is an umbrella term, which can refer both to external negation and to wide scope internal negation, and which most probably can be associated with different syntactic positions. However, the nature of constituent negation has remained relatively undiscussed up until now. The only thing we know on the basis of the tests is that constituent negation is

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



what sentence negation is not. Penka defines non-sentence negation as an event that does not outscope the existential quantifier that binds the event expressed by the main predicate marry, as in (), where the negation not outscopes only the predicate unattractive and not the event in married. ()

He married a not unattractive girl.

On the basis of the Klima-tests it is not clear how to distinguish between affixal negation like un- and low scope negation with not as in () or (). It seems that all different low scope negators are treated as one big bunch, with any differences between them irrelevant to syntax and—from a formal semantic point of view—irrelevant to semantics. In section . we will provide support from morphology pointing to the significance of all these different negative markers to understand the bigger picture of what it means to be negative.

.. W     

.................................................................................................................................. In his analysis of constituent negation, Klima (: ) says that “while the negative elements in sentence negation and constituent negation are the same, the relationships between the negative elements and the sentence in constituent negation, on the one hand, and sentence negation on the other, are grammatically independent of one another.” For Klima this means that the [neg] in constituent negations like () is the same [neg] as in sentence negation, namely a preverbal [neg], but then one in an embedded relative clause. Even though this may be the right analysis for some instances of negative modification, this is not a viable analysis for, for instance, un-prefixed adjectives like unhappy. In what follows we will have a closer look at the morphology of negative markers in order to come up with a more fine-grained classification than the dichotomies that we started out with. De Clercq’s (, b, , to appear) approach to negative markers does not focus only on sentence negation, but also on those negative markers that can be classified as instances of constituent negation. The key idea is that there is a difference between negation and a negative marker. Negation—being syntactically expressed by means of the abstract feature [neg]—is always packaged with other features in natural language negation. Differences between negative markers arise due to the other features [neg] is packaged with. These features determine the scope, semantics, function, and distributional properties of [neg]. One of the crucial tools in De Clercq’s work is the stacking test. This test has been occasionally used in the literature to check whether negative markers are in complementary distribution, as for instance in Chung () for Korean. Whilst Horn’s (a) work went a long way showing that the stacking of different negations always has semantic and pragmatic import, thus criticizing the Stoic idea that negations can be iteratively stacked without any semantic import apart from cancelling each other out, De Clercq’s work shows

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



  

that stacking is not free in the first place, and that there are clear morpho-syntactic constraints related to which negative markers can be stacked. De Clercq (, , to appear) checks negative markers in copular clauses with adjectival predicates in nine languages ( languages in De Clercq, to appear) of diverse origin for their capacity to be stacked upon another negative marker within the same clause. In order to illustrate what is meant with stacking, take for instance the prefix un-. This prefix can occasionally be preceded by non-, but never the other way around, (). ()

a. Nonunhappy people are the best. (p.c. Jim McCloskey) b. *Unnonhappy people are the best.

In the same way, not can stack on un- and non-, but not the other way around, as in (), and n’t can stack on not, but not the other way around, as in (). ()

a. Kim is NOT unhappy. b. Kim is NOT nonprofessional. c. He isn’t not happy.

()

a. *Kim is un-not happy. b. *Kim is non-not professional. c. *He is notn’t happy.

If a negative marker does not tolerate any other markers above it, we can assume that that marker has widest scope, and the marker upon which all other markers can be stacked has lowest scope. As such, the stackability test turns out to be a test into the scopal properties of negative markers, assuming that negation usually takes scope in its surface position. Interestingly, if we look at the functions of these scopally different negative markers, we can see that different functions can be attributed to them. n’t is used for denial (see also Horn a: ch. ), whilst low scope not is used for modifying, (a), or contrasting, (b), the affixal negator non- is used as a classifying negator, (c)–(d) (Kjellmer ), and un-, iN-, dis- are used as characterizing negators, (e) (Kjellmer ). ()

a. a not very happy man, not long ago. b. John was not happy, but sad. c. Nicola believes herself to be a non-angry person and, indeed, she never loses her temper. (Corpus: ukbooks/.) (Kjellmer : –) d. Use non-fat milk instead of whole milk. (Corpus: npr/.) (Kjellmer : –) e. Lainey had a terrible voice, unmusical and sharp, and she usually pitched herself an octave below the sopranos to submerge it. (Corpus: usbooks/.) (Kjellmer : –)

De Clercq (, , , to appear) ends up distinguishing four different types of markers in copular clauses with adjectival predicates. The labels she attributes to the four

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



Table .. Classification of negative markers Tneg-markers

Focneg-marker

Classneg-markers

Qneg-markers

scope over

tensed predicate

untensed predicate

predicate term

predicate term

stack

on Foc, Class, Q

on Class, Q

on Q

-

semantic function

contradiction denying

contradiction contrasting/modifying

contradiction classifying

contrariety characterizing

n’t/not

not

non-

un-/iN-/ dis-, a-

different types of markers correspond to four positions distinguished in clausal syntax. Characterizing markers like un-, dis-, iN- take scope in a dedicated position for negation above a gradable predicate, referred to as QP, a position in the extended projection line of AP (Corver ). Classifying markers like non- take scope in a position above a classifying predicate, referred to as ClassP, a position usually associated with the extended projection line of nominals (Borer ). Contrasting and modifying markers like low scope not take scope in a position above Focus, referred to as FocP, a position associated with the periphery of vP (Belletti , ) or the left periphery of the clause (Rizzi ). Regular verbal negators like not or n’t take scope in a position for negation associated with tense, that is TP (Pollock ). The results of this classification are summarized in Table .. The negative markers in Table . are ordered from left to right according to their scopal properties, that is from wide to narrow scope. Crucial De Clercq’s work is that after a closer study of these negative markers in nine different languages ( languages in De Clercq, to appear), this order—based on the scopal properties of markers—was confirmed by the morphology, and in particular the syncretisms between these different types of negative markers. It turns out that there is no language in which a Tneg-marker is syncretic with a Qneg-marker, unless the Focneg-marker and Classneg-markers are also syncretic. The Czech data in () illustrate this situation of full syncretism (De Clercq : –): (a) illustrates that ne can be used as a Tneg-marker, (b) shows it used as a Focneg-marker and Qneg-marker, (c) illustrates what could be a Classneg-marker and (d) exemplifies the use of ne as a Qneg-marker. () a. Ja ne- jsem st’astný. I neg- am happy ‘I am not happy’ b. On je ne nešt’astný, on je št’astný. he is neg neg- happy, he is happy ‘He is not unhappy, but happy.’ c. Jeho metoda je ne- komerční. his method is neg- commercial ‘His method is noncommercial.’ d. Ja jsem ne- st’astný. I am neg- happy ‘I am unhappy.’

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



  

Table .. Syncretism patterns for negation based on De Clercq (to appear) Tneg

Focneg

Classneg

Qneg

English French (informal) Swedish Turkish Japanese

dhen ne …pas (-ci) an(i) (ha-) (-ci) mos (ha-) not pas inte degil nai

oxi pas an(i) mosnot pas inte degil nai

mi non pimolnon non ickegayri/olmayan hi-

aiNpulmoluniNo-siz hu(/bu)/mu

Khwe

vé

ŋya

ó-

ó

Chinese MS Arabic Persian Mayalayam

bù laa na alla

bù laa na alla

fēi ghayrqheyra-

fēi ghayrqheyra-

Moroccan Arabic

ma (ši)

muši

muši

muši

Hungarian Hebrew Dutch

nem lo niet

nem lo niet

nem lo niet-

-tElEn biltion-

Russian Czech Malagasy Hixkaryana Tümpisa Shoshone

ne netsy -hike(e)

ne ne tsy -hike(e)

ne netsy -hike(e)

nenetsy -hike(e)

Greek French (formal) Korean

Greek (Table .) illustrates how all types of negative markers can also get a different lexicalization, i.e Greek is fully non-syncretic. Table . illustrates the syncretism patterns, as shown by the shading. It also reveals that there are no ABA-patterns, thus respecting the well-known descriptive *ABAgeneralization. The *ABA-generalization is a restriction on patterns in a paradigm, which says that it is possible to order paradigms in such a way that only contiguous cells in a paradigm are syncretic (Baunaz and Lander ; Caha and Vanden Wyngaerd ). The ordering of the paradigm in Table . confirms the scope of negation, that is, the order from wide to narrow scope in Table ., thus strongly suggesting that morphology is not arbitrary. Concretely, of the fifteen logically possible patterns shown in Table ., those shown in gray are ABA patterns, i.e. patterns showing a syncretism in noncontiguous cells across a nonsyncretic form, and they are not attested.² All those that are attested (cf. Table .) are not ABA patterns.³ ² The table in . is based on Michal Starke’s summary of the patterns in my work. He was so kind as to share this with me. All mistakes are of course mine. ³ De Clercq (to appear) argues that there is at first sight a missing pattern, that is pattern  in Table . (Table . shows seven attested patterns and there are eight logically possible patterns). This can be due

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



Table .. Logically possible patterns 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

A A A A A A A A A A A A A A A

A A A A A B B B B B B B B B B

A A B B B B B B A A A C C C C

A B A B C A B C A B C A B C D

Based on the patterns in Table ., and in particular the languages in the bottom row, it is clear that all negative markers share something, a meaning that we call negation. However, what the patterns also show is that these markers are also somehow different. This is very well illustrated by Greek, where a different negator is associated with all four types of markers. As such, the system that we need to set up should ideally capture both the similarity and the differences. In order to capture the similarity and differences between negative markers, negation can be decomposed into its subcomponents (Caha ; Starke ). The internal syntax of negative markers needs to consist of at least one identical feature. The obvious candidate here is [neg]. The distributional and scopal differences are then taken as meaningful reflections of featural differences amongst negative markers, which should be part of their internal syntax. Negative markers like un-, (c), which take scope above a gradable predicate, that is QP, consist in addition to a [neg] feature also of a [Q] feature. Negative markers like non, (b), which take scope above a classifying predicate, that is ClassP, add a [Class] feature, and Tneg markers add a T-feature. The internal structure of the different negative markers in English is shown in (). The difference in internal structure will determine where the different markers can appear.⁴

to the limited set of languages studies. However, upon closer examination, the missing pattern  can be found in a variety of French. It is beyond the scope of this chapter to go deeper into this and I refer the reader to De Clercq (to appear) for more discussion. ⁴ De Clercq’s system is couched in Nanosyntax. Phrasal spellout, the Superset Principle, and the Elsewhere Principle govern lexical insertion in Nanosyntax and allow us to capture the *ABA-pattern. It is beyond the scope of this chapter to discuss spellout in detail and how the different markers project in clausal syntax. I refer the reader to Starke (, ), Caha (), and Caha, De Clercq, and Vanden Wyngaerd () for more discussion of the nanosyntactic programme and to De Clercq () for a concise overview of the nanosyntax of negation.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



  

()

a. [TP T [FocP Foc [ClassP Class [QP Q Neg]]]] ⇒ not b. [ClassP Class [QP Q Neg ]] ⇒ non c. [QP Q Neg] ⇒ un-

The take-away message from this discussion is that the basic dichotomy between sentence negation and constituent negation is too coarse to do justice to the realm of negative markers the languages of the world have at their disposal. On the basis of a cross-linguistic investigation of syncretism patterns, it can be argued that at least four different types of negative markers can be distinguished, three of which are candidates to resort to the label ‘constituent negation’, that is Focneg-markers, Classneg-markers, and Qneg-markers.

.. T  

.................................................................................................................................. A question that emerges on the basis of the discussion in section . is how the negative markers in De Clercq’s classification relate to the dichotomies we started out with. The Tneg-markers in De Clercq’s system are those that normally give rise to sentence negation according to the Klima-tests, unless other operators co-occur with them. They are the regular verbal negators a language has at its disposal (the standard negators in Payne  and Miestamo’s  terminology) and this group is internally complex, consisting of various different morphological types (particles, negative verbs, affixes, cf. Dryer ). However, it is precisely due to the fact that other operators can occur in a structurally higher position and then change the polarity of the clause, (b), that it seems necessary to distinguish another position and possibly another type of negative marker, responsible for sentential polarity, which is structurally higher than the one for regular sentential negation. The assumption is that it is this higher position that the Klima-tests, like the question tag test, are sensitive to. If there is no other operator available, the Tneg-marker can determine the polarity of this high position (via Agree, overt or covert movement). However, if there is another operator in between the Tneg-marker and this high position, the polarity of the clause will be determined by the polarity of this higher operator. Our discussion of this high position brings us back to the discussion of external negation, and more in particular to the external negator lāw identified in Jewish Babylonean Aramaic (JBA) by Bar-Asher Siegal (a, b). Even though it is indeed generally assumed that the notion of external negation is one that belongs to the realm of logic or formal semantics rather than to the realm of natural language (cf. Horn a: ), it seems that there is empirical evidence in the form of a negative marker after all. Bar-Asher Siegal and De Clercq () extend the idea developed on the basis of JBA to the negator neca in Sicilian (Garzonio and Poletto b), arguing that the external negators in JBA and Sicilian are linked to a position in the left periphery of the clause, that is, to a left peripheral Focus Position, FocP, thus linking external negation to the position that was also identified as a position for sentential polarity. Even though it seems that there are not many languages with a dedicated marker for external negation, all languages have strategies to express external negation, that is by means of negative clefts, negative preposed constituents, negative quantifier subjects, negative polarity particles, or real external negators. As such, it is not surprising that this

Table .. Terminological overview  JBA law/Sicilian neca Preposed PP/DP negative clefts¬

n’t/not

external negation Klima Frege

not

non-

un-/iN-/ dis-, a-

internal negation sentence negation

constituent negation

predicate denial

predicate term negation

propositional negation

Aristotle/Horn

affixal/lexical negation

Payne/Miestamo

standard negation

Jespersen

nexal negation

special negation

sentential negative polarity syntactic classification

(FocP)

Tneg-markers

Focneg-markers

Classneg-markers

Qneg-markers

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

Zimmer/Dahl

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



  

left peripheral position, FocP, has been argued to host focal constituents and wh-words in Italian (Rizzi ), negative preposed constituents (Haegeman ), and polarity particles like yes/no (Holmberg ). It is a position dedicated to negative and interrogative clause typing and hence can be associated with sentential polarity and external negation. It should not be a surprise that many languages use the Focneg-marker also as a polarity particle, meaning ‘no’, suggesting interaction between the low scope position for Focnegmarkers and the left peripheral FocP. Summarizing, sentence negation is an umbrella term referring to different types of wide scope negation that are interpreted externally or internally. In the latter case the negative marker does not succeed in making the polarity of the clause negative due to the presence of another operator, as discussed for the example in (b). With respect to constituent negation we can say that all constituent negation is internal negation, but then with different scopes. De Clercq’s Focneg-marker, is the constituent negator with widest scope. It can best be classified under the Aristotelian concept of predicate term negation, just like Classneg- and Qneg-markers. Table . is an update of Table .: it shows how the terms discussed in this chapter relate to each other and it aims at providing an overview of and insight into the terminological minefield surrounding negation.

.. C

.................................................................................................................................. What we have established in this chapter is that there is a difference between two dichotomies: (i) external negation vs. internal negation, and (ii) sentence negation vs. constituent negation. We discussed the Klima-tests, how they relate to the sentenceconstituent negation dichotomy, and we reviewed some of the criticism the tests received in the literature. We argued that the term ‘sentence negation’ is an umbrella term, referring at least to external negation and wide scope internal negation. We established that the same is true for ‘constituent negation’, that is it is too broad a term to do justice to the different types of negation the label covers. In line with this idea, we discussed the relevance of syncretisms between various types of negative markers, which led to a four-way classification of markers, ordered according to their scopal type. Finally, we discussed how external negation is a term confined to that type of negation which takes scope over the entire proposition, including the subject, and we mentioned that it has been argued that there is no specific morphological instantiation of an external negator, a claim which has recently been problematized.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

  ......................................................................................................................

  ......................................................................................................................

 

.. I

.................................................................................................................................. N is one of those phenomena that have captured the interest of scholars from diverse fields like philosophy, linguistics, psychology, and sociology. It is usually studied as a phenomenon in logic, where it is viewed as a play of contradictories and contraries with their shades of truth values and falsities. It is also an integral part of human natural language, differentiating the latter from other modes of communication, especially animal language, as discussed by Horn (a) in his introduction to his book. In this chapter, we will study affixal negation—negation brought about by affixation—in language. Most studies on affixal negation elaborate on the affixes used for negation. The current chapter, however will focus on the phenomenon of negation, rather than having the affixes for negation at the center of attention. The two are closely tied to each other; even so, having the focus on one or another, alters the perspective slightly. The majority of examples in this chapter are taken from English. It must, however, be noted that other languages like Sanskrit, French, German, Marathi (among others) also exhibit the phenomenon of affixal negation. So, it is not a phenomenon restricted to English, but can be observed in other known languages, as well. In this chapter, we will start by taking a cursory look (instead of a detailed one, for reasons of space and scope) at the various types of negations carried out in language. After determining the place of affixal negation in this broad spectrum of negation, we will examine the characteristics of affixal negation vis-à-vis other types of negation and how it differs from the analysis of other types of negation. Next, we will investigate how affixal negation is brought about—the morphological and semantic structures and processes involved in it, and also elaborate on the nature of the linguistic elements involved in the process. Affixal negation is not homogeneous. One can visualize various categories therein, based on the kind of semantic information conveyed. We will discuss in detail how affixal negation can be categorized. The chapter will end with some concluding remarks.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

.. A      

.................................................................................................................................. Simply put, negation in language is the verbal act of denial or refusal. There are, of course many more shades to this act and there are also various linguistic instruments present in all known human languages which help to bring these shades into effect. It is believed, however, that this simple act of negation carries beneath its surface different dimensions of human behavior. Consequently, it has been a subject of study over the centuries and for scholars from diverse fields, like logic, philosophy, psychology, sociology, and, of course, linguistics. A detailed discussion of all these perspectives on negation is out of the scope of this chapter due to the constraints of space, but this section will try to examine some of the more relevant perspectives and try to establish a place for affixal negation in this spectrum of negation.

... Negation in philosophy and logic Some of the earliest thoughts on negation come from the Indian and Greek philosophical traditions. The discussions on negation in Mīmāmsā and Nyāya in the Indian philosophical _ ‘present’ vs. the ‘absent’ and on the tradition are enquiries into the notions of the perception of absence. Coming from the Greek philosophical tradition, the Aristotelian square of negation, depicting the notions of contradictory and contrary negation (and the subcontraries), has inspired a lot of work on the subject. The difference in perspectives on negation between logic and linguistics can be illustrated with the help of the following sentences: ()

a. The king of France is bald. b. The king of France is not bald.

Sentence (a), when uttered in today’s world, has different implications in logic. Its analysis in logic is ambiguous. For some logicians, it is a meaningless sentence since it can be neither  nor . For others, it is simply . This is due to the fact that today’s France has no king. As a consequence, its negation—sentence (b)—too is problematic. Sentence (a), however, is an acceptable sentence in language—in English (be it in a real scenario, or a fantastic one). It can be accepted and analyzed phonologically, morphologically, syntactically, and semantically. All the phonemes in it are phonemes belonging to English. All the morphemes contained in it are meaningful and acceptable in English. Its construction follows the syntactic norms of English. Semantically, its meaning can be understood very well. Further, the second sentence can be said to be a negation of the first sentence and it too is an acceptable sentence, linguistically speaking. These notions of contradictory and contrary negation hold true in linguistic negation as well, but they tend to be expressed differently in language than in logic. Consider the following sentences (examples taken from Horn a):

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

  ()

a. b. c. d.



Every man is white. No man is white. Not every man is white. Some men are not white.

In the Aristotelian square of opposition, the sentence (a) has sentence (b) as a contrary opposite and sentences (c) and (d) as contradictory opposites. Depending upon which aspect of the meaning of the sentence is to be negated it can have more than one opposite. Negation of sentence (a) based upon quantity would be (b) and (c) Negation of sentence (a) based upon quality would be ()

Every man is not white.

Sentence () may imply sentence (d), but the latter one does not come forth as an obvious negation of sentence (a) based upon quantity. While these two types of negation in logic focus on the truth or falsity of propositions, linguistic negation is far more varied (and at times even more subtle), resulting in various shades of negation in addition to those expressed as contradictions and contrarieties. In affixal negation, for example, one can express negation not only in terms of direct refutal or denial of the existence of the whole or a part of the meaning of the base, but also consider other possibilities to express negation, in terms of reversal, deprivation, qualitative and quantitative judgments, and so on. Section . in this chapter discusses in detail these aspects of the semantics of affixal negation.

... Implicit or pragmatic negation It is observed that negation is not always expressed in explicit terms. Negation can be brought about implicitly, through a combination of some linguistic or prosodic means (like tone, accent, etc.) and some extra-linguistic means (like bodily gestures, contextual interpretation etc.). Sarcasm is a means of bringing about such implicit kinds of negation in language. Consider the following example: ()

You’re a great help, thank you!

In this example, there is no apparent explicit negation. However, uttered in a certain way, punctuated with sarcastic tones, with accentuation of certain words, possibly in combination with certain bodily gestures and in corresponding context or situation, this sentence actually might mean ‘you have not been of any help at all!’ It thus conveys a negative meaning, implicitly. Another aspect of speech that results in an implicit negation is the intonation and (emotive) accentuation. Those elements are stressed upon or accentuated that need to be focused upon in the utterance. In this case, the sentence construction need not be varied, but the accent on certain words may shift. As is apparent from the following examples, the negations of various elements in the sentence () are effected using

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

appropriate accentuation on the corresponding words (examples (a), (b), and (c), with the accentuated words shown in bold). ()

John ate an apple.

()

a. Peter ate an apple. (= not John) b. John ate an apple. (= not cut) c. John ate an apple. (= not pear)

Thus, it can be seen that accentuation (along with appearing in appropriate context) helps in bringing about an implicit type of negation.

... Syntactic negation Linguistic negation can be categorized into syntactic (or sentential) negation and morphological negation. Syntactic or sentential negation is more commonplace and also gets a lot of attention in research as compared to morphological negation. (See Jespersen ; Horn a; Geurts .) Syntactic negation distinguishes itself from morphological negation (of which affixal negation is a part), primarily by the end result of creation (or not) of a new lexeme. Affixal negation, being a derivational process, results in the creation of a new lexeme, whereas syntactic negation does not. Sentential negation is brought about by using negating elements like ‘not’, ‘neither’, ‘none’ etc. (in English). Typically, these elements occur tied to the element that they are negating in the sentence and have a scope of negation over these adjacent elements in the chain. This scope can, however, stretch beyond these immediate neighboring elements. Thus, there is a possibility of saying ‘This is really not the most desirable of outcomes,’ with the scope of the negating element ‘not’ extending to the whole unit ‘the most desirable of outcomes’ and not being limited to its immediate neighbors—‘the’ or ‘the most’. Second, the scope of sentential negation is elastic. Thanks to the relative syntactic autonomy that these negating elements enjoy, this scope can be extended by allowing a certain flexibility in their placement in the sentence. For example, the negating element ‘not’ can be placed at different positions to indicate the negation of different elements in the sentence. Consider the following example: ()

John ate an apple.

There is a possibility to negate more than one element in this sentence. Depending upon which element is to be negated and what more information (if any) is to be given out, the position of the negating element ‘not’ can be altered appropriately. Thus, we have different negations of this same sentence (), as shown in the following sentences (a), (b), (c), and (d). ()

a. b. c. d.

John did not eat an apple. John ate a pear, not an apple. John did cut, but not eat, an apple. Peter, not John, ate an apple.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

 



... Morphological negation Morphological negation, as opposed to sentential negation, can be defined as a type of negation where the negation is brought about by modification of the morphological structure of the involved elements. So, different morphological phenomena may be involved in this process of negation. Multiple ways of bringing about morphological negation are discussed by Cartoni and Lefer (): Affixation - Prefixation: Attaching a prefix to a base to form a negative derivative (e.g. unselfish, dishonest) - Suffixation: Attaching a suffix to a base to form a negative derivative (e.g. cashless) Compounding: Two free forms coming together to form a negative derivative (e.g. hands-free) Conversion: Forming a negative derivative without altering the form of the base (e.g. to dust = to remove dust, to shell = to remove a shell) In addition to these morphological processes, the role of reduplication as a phenomenon contributing to negation has also been studied. Bond () elaborates on negating reduplication in Eleme, a language spoken in Africa. Miyake () has studied the formation of negative derivatives through reduplication in Javanese. Thus, affixal negation is explicit morphological negation. The sections that follow will explore different aspects of this phenomenon.

.. M   

.................................................................................................................................. In the structural perspective, affixal negation is negation carried out with the help of an affix attached to a base. This affix can be ‘negative’ or ‘positive’. Some affixes are considered to be ‘negative’ based on the fact that the majority, if not all, of their derivatives are negations (like iN-, anti-, -less, etc.). It must be noted, though, that some of the affixes considered as ‘positive’ (or at least not considered as ‘negative’) also bring about affixal negation (like ‘hyper-’ in ‘hyperactive’). So, while studying affixal negation, it is better to focus on the process, rather than on any particular affix, by labelling the affixes as ‘positive’ or ‘negative’. Affixal negation helps in avoiding the construction of entirely ‘negative’ phrases. So instead of saying ‘John is not cooperative’ one can say ‘John is uncooperative’, which on the whole appears to be a more positive phrase, as compared to the first one. Affixal negation also allows us to do double negation in a sentence easily. So instead of saying ‘The act was intentional’, we can say ‘The act was not unintentional’. Such a technique comes in particularly handy in various tricky situations in socio-cultural interactions where it is better to be subtle and to circumvent or paraphrase a direct proposition in order to avoid confrontation.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

... Structure of affixal negation Affixal negation is derivational in nature—it creates a new lexeme. So, the elements participating in the process of affixal negation are base and affix. This affix can be a prefix or a suffix. ()

a. in- + direct = indirect b. dis- + arm = disarm c. cash + -less = cashless

It is observed that the process of suffixation is much more prevalent in language than the processes of prefixation and infixation, as discussed by Sapir () and Cysouw (). Although these observations do not make an explicit distinction between inflectional and derivational affixes, and the majority of the examples of affixation that are discussed in these works are inflectional in nature, Sapir also mentions a few examples of derivational affixation. It is, of course, extremely difficult to quantify these observations with certainty (due to a number of factors like ambiguity over a clear definition of a prefix and a suffix, appropriate corpora, etc.) and a discussion of them would be out of scope of this chapter. In the particular case of affixal negation, however, this observation (of suffixes outnumbering prefixes) does not seem to hold true. Negation by prefixation is observed to be much more abundant than negation by suffixation. As a consequence, the number of prefixes that bring about negation seems to be larger than the number of suffixes that bring about negation.

... Grammatical category of the base Affixal negation is effected on bases belonging to different grammatical categories like nouns, adjectives, and verbs. Out of these, bases belonging to any of these grammatical categories may undergo affixal negation of the types contrary and contradictory (or direct, as discussed in section .). So, we have the negative derivatives of nouns (mercy– merciless), of adjectives (logical–illogical) and of verbs (fire–misfire). Further, the same affix may bring about negation in bases belonging to different grammatical categories. For example, the negating prefix ‘dis-’ will form ‘disagree’ from the verb ‘agree’ and ‘discomfort’ from the noun ‘comfort’. However, for bringing about negation of other shades (or indirect negation, as will be seen in section .), only those types of bases that display certain pertinent characteristics, will be brought into use. Thus, for the negation of the type of ‘reversibility of action’, only those bases that indicate an action that can be reversed will be brought into use. For example, the prefix un- (in its ‘indirect’ mode) will attach itself to the verbal base ‘tie’ to give ‘untie’. In this respect, it is unlikely that it will attach itself to a noun or an adjective or even a verb that does not show this property of ‘reversibility’, like ‘marry’. Hence the negative derivative *‘unmarry’ does not exist. (It must be noted here that the derivative ‘unmarried’ exists, but it is a ‘direct’ negation of ‘married’ and not an ‘indirect’ negation with the shade of ‘reversibility’. A detailed discussion of this distinction between ‘direct’ and ‘indirect’ negation can be found under the semantics of negation in section ..)

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

 



... Grammatical category of the derived word Since affixal negation can be considered as a type of affixal derivation, the norms of derivation regarding the grammatical categories of base and the derivative—whether there is a difference or not between them—can also be said to apply to affixal negation. Thus, if affixal negation is carried out by prefixation, the grammatical category of the derived lexeme is usually the same as that of the base. So, in ‘paid–unpaid’, the grammatical category of the base and the derived lexeme is the same—an adjective, in ‘arm–disarm’, the grammatical category remains the same—a verb—while deriving the negated lexeme. There are, however, some instances of negation by prefixation altering the grammatical category between the base and the derivative. For example, in deriving ‘anti-tank’ from ‘tank’, the grammatical category has gone from a noun to an adjective. In the case of affixal negation being carried out by suffixation, the grammatical category of the derived lexeme may be different from that of the base. In the case of ‘cash–cashless’, the grammatical category of the base is a noun and that of the derived lexeme is an adjective. Similarly, in ‘figure–disfigure’, one can see the change in the grammatical category from a noun to a verb.

... Scope In the case of affixal negation, the affix negates the immediate element (in whole or in part, semantically speaking)—which is the base to which it is attached. There is little elasticity or flexibility in a structural perspective, unlike the one for sentential negation. It is to be noted here that one can observe constructions like ‘un-bloody-believable’ where an infix ‘bloody’ is inserted between the prefix un- and the base ‘believable’. Some more examples of such infixes are ‘friggin(g)’ and ‘fuckin(g)’. Taken out of this context, they usually have a pejorative value. In such cases, the infix is just an emotive intensifier and is neither the real base to be negated nor a negating affix. Even if the negating affix (‘un-’, in this case) seems to be attached directly to this intensifier infix, it really negates the base, which is positionally farther. Hence, the scope of the negating affix seems to stretch beyond its immediate neighboring element. These are, however, considered as exceptional cases in English.

... Co-occurrence constraints Zimmer () observed that negative affixes cannot be used with (adjective) bases that have a ‘negative’ value. Thus, we do not have ‘*unsad’ or ‘*unfalse’. Horn (a) further expanded on the nature of the ‘positive’ value of the base, stating that negative affixes can attach to a base that is unmarked and that has a weak positive scalar value. Thus, we observe derivatives like ‘dislike’ exist, but not ‘*dislove’, where ‘to love’ is considered to be a stronger sentiment than ‘to like’ and thus has a stronger semantic value. De Clercq and Wyngaerd () built on the arguments by Zimmer and Horn to show that this restriction applies to co-occurrence of more than one negative affix in the same derivative, preventing them from occurring in the same functional sequence. Hence, one can observe that negative derivatives like ‘*ununhappy’, ‘*unuseless’ are considered as invalid.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

This argument can be further expanded to include those affixes that are not generally considered as ‘negative’ but do bring about negation, such as ‘hyper-’. So, the negative derivatives of such affixes too do not accept any more negative (or negating) affixes (for example, *unsubnormal).

.. S   

.................................................................................................................................. Although structurally the affix is attached to the base, affixal negation may work on the whole or on some part of the meaning of the base. Negation in many cases is characterized by the markedness with respect to an unmarked state of meaning. It is observed that in a pair of antonyms, the unmarked element is taken to be the ‘positive’ (or at least ‘neutral’) element and the marked one is taken to be the ‘negative’ one (Lehrer ). Similar behavior is observed in the derivatives produced by affixal negation, as discussed in Joshi (). Within the pair ‘happy–unhappy’, identifying the positive and negative elements is easy. However, in the case of derivatives like ‘normal– subnormal’, this notion of markedness comes into play. Thus, in this pair, the unmarked element ‘normal’ is taken to be a positive and the marked element ‘subnormal’ is taken to be negative.

... Negation of the denotative or connotative meaning Affixal negation may work on the denotative (or the explicit or the ‘observable’) meaning of the base or on the connotative (or the implicit or the ‘hidden’) meaning of the base. This is illustrated by the affixal negations carried out by the prefixes ‘non-’ and ‘un-’. The prefix ‘non-’ produces negations like non-American, non-Catholic, etc. The prefix ‘un-’ may also produce negations by attaching itself to the same bases, for example, unAmerican, un-Catholic, etc. The difference in these two cases of negation lies, however, in the semantic content that is being negated. A non-American person is a person who is not American by nationality. So, it can be anybody other than an American. An un-American way (of living, for example) is one which does not follow the norms of living (social, cultural, economic, etc.) followed in American society. This may apply to persons who may or may not be American by nationality. So, the first kind of negation (the one with ‘non-’) targets the explicit meaning of being American (i.e. possessing the said nationality). The second kind of negation (the one with ‘un-’) targets the implicit meaning that is attached to being American (i.e. following the norms (social, cultural, economic, etc.) regarded to be typically American).

... Which aspect of the meaning of the base is negated? As discussed earlier in this section, the negation in affixal negation may work on denotative or connotative meaning of the base. In either case, there would be some particular aspects

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

 



of the meaning of the base that would be acted upon by the process of negation. Some such aspects are: Spatio-temporality: Negation is demonstrated in the form of inappropriateness of space or time for a particular action. Directionality: Negation is demonstrated in the form of reversal of direction, going against something/somebody etc. Usually the ‘positive’ is the word indicating the direction of going forward, while the negation is expressed by the direction backwards or in the opposite direction or against the ‘positive’ counterpart. Reversibility: If there exists a semantic component in the meaning of the base to indicate whether the action/event indicated by the base can be reversed, it may be subjected to negation. The negation in this case is usually indicated by the ‘reversal’ of the action/event indicated by the base. Qualitative judgment: Whether there is a possibility of passing a judgment of good vs. bad, desirable vs. undesirable, above or at par with or below expectations, etc. The ‘positive’ element, in this case, is the unmarked or the one meeting normal/accepted expectations. The negation is indicated by a deviation from the normal. Quantitative judgment: Whether there is a possibility of passing a judgment to the effect of overabundance, or insufficiency, or lack, or absence of certain traits/elements. In this case too, the ‘positive’ aspect is the one that meets the norms or expectations and any deviation from this norm is the negation. The neologism ‘to unfriend’ presents an interesting case of a base acquiring a new semantic aspect in order to facilitate its negation. This term was coined primarily due to the emergence of social networking in the early s. The term indicates the process of removing someone from one’s follower (i.e. friend) list. In order to do this, one should have first added that person to one’s friend list. The verb ‘to befriend’ has existed in English for longer. However, a new verb ‘to friend’ came into being to show this process of adding someone to one’s friend list. Going further, it acquired the aspect of reversibility to facilitate the reversal of this action of ‘friending’ someone, that is the removal of someone from one’s friend list. A detailed discussion of categorization of affixal negation based on the negation of these aspects of the base will appear in section ..

... Affixal negation of ‘neutral’ Even though affixal negation is a type of negation, it may not always negate a ‘positive’. One can observe cases of affixal negation where the negation is with respect to a ‘neutral’ term rather than that of a ‘positive’ term. Consider the examples below: ()

conseiller vs. déconseiller

French

()

active vs. hyperactive

English

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

The word ‘conseiller’ means ‘to advise’, which indicates just an action without any explicit or implicit positive connotation to its meaning. The negation, however, means ‘to advise (somebody) to not (do) something’ or ‘to advise against (doing something)’. Now, this is a negation which does not have any opposite ‘positive’, explicit or implicit. Another case is that of ‘active vs. hyperactive’. The affix ‘hyper-’ does not negate the state of being active; it negates, instead, the state of activeness considered as ‘normal’. A hyperactive person is still active, but to an extent that is more than required or more than ‘normal’. In this case, there is no explicit ‘positive’ of the term ‘active’. One can argue that the meaning of ‘active’ contains the implicit positive. Thus, there is no need of an explicit positive. So, what the affix ‘hyper-’ negates is that implicit positive. In such cases, it can be said that the process of affixal negation is negating a ‘neutral’, rather than a ‘positive’.

... Affixal negation of ‘positive’ Sometimes, affixal negation brings about a complete change in polarity from positive to negative. For example, in the pair ‘careful–careless’, one element (‘careful’) is positive and the other one (‘careless’) is negative. Although morphologically both of them are derivatives of ‘care’, none of them are connected to it if we consider the relation of opposition. Thus, even if ‘careful’ is positive, ‘care’ is not its negative and even if ‘careless’ is negative, ‘care’ is not its positive. Thus, in the pair of opposites, a middle ‘neutral’ element does not exist. A curious case is presented by the pair ‘gruntled–disgruntled’. According to the Oxford English Dictionary, ‘disgruntled’ was coined in English circa the seventeenth century. It did not exist as a negation of ‘gruntled’ because the latter did not exist in English at the time. However, ‘gruntled’ came to be coined in English in the s, possibly due to the misperception of ‘dis-’ as a negating prefix, through back-formation. Thus, a positive term was created to correspond to a negative term.

... Affixal negation with no ‘positive’ or ‘neutral’ counterpart Sometimes, it is observed that there is no real ‘positive’ counterpart to a negative element. The case of ‘gut vs. gutless’ can be cited as an example of this. First, the word ‘gut’ is a substantive without any ‘positive’ or ‘negative’ shades to its meanings, as opposed to an adjective. Second, there are no explicit positives (either in the base form or in a derivative form) of the word. So, the affixal negative derivative ‘gutless’ does not negate any explicit positive forms. It must be noted here that ‘gutful’ exists but it is not the opposite of gutless. Besides it is a less common variant of bellyful. Similar observation can be made for the pair ‘moralize’ vs. ‘demoralize’. The word ‘moralize’ has a meaning that is different from what a ‘positive’ of ‘demoralize’ would have. Thus, not only is ‘demoralize’ not a negation of ‘moralize’, it has no real positive counterpart.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

 



.. C   

.................................................................................................................................. Over the years, there have been a few attempts at categorization of negation. Some of them apply to affixal negation as well. For example, the classic Aristotelian categorization of ‘Contrary’ and ‘Contradictory’ negation applies to affixal negation, as discussed by Jespersen (), Zimmer (), and Horn (). There can be multiple ways of categorizing affixal negation, based on various criteria—structural and semantic. Structural categorization of affixal negation considers the structure/construction of the derivative and its constituents in terms of their structural and syntactic properties. Semantic categorization of affixal negation is more complex. It considers the semantics and pragmatics of the derivative. So, the negation may or may not be obvious at the outset.

... Structural categorization of affixal negation The first and obvious way of categorization would be positional, that is based on the position of the affix with respect to the base. Thus, we can categorize the derivatives based on whether the negating affix is a prefix, a suffix, an infix, or any other. So, the categories would be like prefixal, suffixal, infixal, etc. For example, the derivative ‘discontinue’ would be prefixal, the derivative ‘careless’ would be suffixal. Another type of categorization would be the one based on the grammatical category of the negating affix. If we label the affix with the category of the source and the target lexemes, we will have a nomenclature like ‘denominal adjectival’—an affix that attaches itself to a nominal (i.e. substantive) base and the derived lexeme is an adjective. So, one could imagine a two-tier categorization with the first tier with the grammatical category of the derivative and the next tier would denote the grammatical category of the source lexeme or the base. Thus, the first tier would contain categories like verbal, nominal, adjectival, adverbial, etc. The second tier would contain categories like denominal, deverbal, deadjectival, etc. The whole description would be a combination of the two tiers—verbal denominal, adverbial deadjectival, etc. For example, the derivative ‘undo’ would be categorized as ‘verbal deverbal’, the derivative ‘anti-terrorism’ would be ‘adjectival denominal’, etc.

... Semantic categorization of affixal negation Some researchers (like Cartoni and Lefer ) have defined five categories of (affixal) negation as follows: - Contradictory negation: polar opposites with no middle term (e.g. ‘American’ vs. ‘nonAmerican’) - Contrary negation: scalar opposites with possible intermediate degrees of opposition (e.g. ‘happy’ vs. ‘unhappy’) - Privative negation: negation indicating the lack of an entity (e.g. ‘order’ vs. ‘disorder’)

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

- Reversal: negation indicating a return to original state, which acts on verbal bases (e.g. ‘do’ vs. ‘undo’) - Removal: negation indicating removal of an entity, which usually acts on nominal bases (e.g. ‘arm’ vs. ‘disarm’) - Opposition: negation indicating the antagonistic relation between the base and the derivative (e.g. ‘matter’ vs. ‘antimatter’) Based on the negated element in the process of derivation, affixal negation can be categorized as being ‘direct’ or ‘indirect’, as discussed by Joshi (). The basis of this categorization is the check whether the base is negated with a NOT or by effecting a negation of some other implicit semantic component of the base. The main difference between ‘direct’ negation and ‘indirect’ negation is that in the former one, the existence of the meaning or the entity indicated by the base is refuted. In the latter type of negation, the existence of the meaning indicated by the base is not denied but some other semantic component in its meaning may be shown in a ‘negative’ color. For example, the negation ‘happy vs. unhappy’ is a direct negation and the negation ‘famous vs. infamous’ is an indirect negation. In the first example, the negation refutes the particular state of mind (known as ‘being happy’), which puts it in direct opposition with the original ‘positive’ entity. In the second example, being ‘infamous’ does not negate the existence of ‘famous’, rather it brings out its undesirable side. In other words, being ‘infamous’ is not being ‘not famous’, rather being ‘famous for undesirable reasons’. Indirect affixal negation has further subcategories, based on the various types of meanings or shades of meaning carried by the bases and the affixes, as illustrated below: - Reversal of direction: negation is carried out without negating the concept of movement indicated by the base (e.g. attack–counter-attack). - Reversal of action: negation by indicating an action performed to reverse another previous action (e.g. tie–untie). - Inferiority: negation carried out by qualitative judgment about the hierarchical placement (lower than the one considered optimal or normal) of the element indicated by the base (e.g. tension–hypotension). - Insufficiency: negation through the quantitative judgment about the level, taken as negative in some contexts (e.g. normal–subnormal) - Badness / wrongness: negation in the form of the qualitative judgment of an entity or some of its aspects being unacceptable (e.g. behave–misbehave). - Over-abundance: negation in the form of existence in excessive and undesired quantity of an entity (e.g. active–hyperactive—typically in the case of a child (medically taken to be a disorder)). - Pejorative: negation by qualitative judgment (e.g. drunk–drunkard—indication of excessive drinking). - Opposition: negation by indication of opposition in notion, action, ideology, etc. (e.g. matter–antimatter). - Removal: negation indicating the removal of something (e.g. bug–debug).

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

 



It is observed that these two categories of affixal negation—direct and indirect—are not mutually exclusive. Thus, one affix may be a part of either or both of these categories, depending upon the negation that it displays. For example, the different negative derivatives of the prefix ‘mis-’ make it a member of both the categories. In the derivative ‘misfire’, the negation is of the type direct (to misfire = to NOT fire) and in the derivative ‘misbehave’, the negation is of the type indirect (to misbehave = to behave in an inappropriate manner). Further, even within indirect negation, the derivatives of the same affix may fall under different subcategories, depending upon the polysemy of the affix. So, the shades of meaning carried by the prefix ‘counter-’ in ‘counter-attack’ and ‘counter-productive’ are slightly different. In the first one, it indicates the reversal of direction of attack. In the second one, the negation is not just a reversal, but something that actually is a hindrance to being productive. So, the first negation would fall under the subcategory ‘reversal of direction’, whereas the second one would fall under the subcategory ‘pejorative’. Hence, for any negation or a negative derivative, it is hard to predict a priori the membership of these categories, without a detailed examination.

.. G 

.................................................................................................................................. There have been attempts at taking the categorization of affixal negation and expanding it or using it further for other applications. Categorization of affixal negation is also proven to be helpful for computational purposes. An electronic dictionary of affixal negation, largely based on the categorization described by Joshi (), has been successfully implemented by van Son, van Miltenburg, and Morante (). A good categorization, which is essentially based upon different distinguishing traits (intrinsic or extrinsic) of the entity in question, would facilitate feature engineering, which in turn is the backbone of a machine learning based system for natural language processing. In the case of negation and affixal negation, such a system would turn out to be essential for applications like automatic sentiment analysis, automatic translation, natural language understanding, or natural language generation.

.. C 

.................................................................................................................................. Affixal negation has been a very interesting and complex but relatively underexplored and understudied subdomain of linguistic negation. The prevalent notions of contradictory negation and contrary negation along with the subcontraries help in explaining affixal negation only partially and leave much of it out. Studying affixal negation requires inputs from not only logic, but also from morphology, syntax, and lexicology, since it is a part of linguistic negation.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

Affixal negation differentiates itself from sentential negation in terms of the scope of negation. It is derivational in nature, since the result is a new lexeme. The morpho-syntactic and semantic properties of the bases and affixes play a big role in the process. Further exploration of the phenomenon of affixal negation would certainly bring out some more aspects and would also help in refinement of categorization. Besides adding to the linguistic knowledge base, such a refinement in categorization would also be helpful in developing various computational systems for natural language processing.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

  ........................................................................................................................

QUESTIONS IN THE SYNTAX OF NEGATION ........................................................................................................................

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

  ......................................................................................................................

    ......................................................................................................................

      

.. I

.................................................................................................................................. S Payne () and especially Miestamo () typologists have used the terms ‘standard’ and ‘non-standard’ negation.¹ ‘Standard negation’ is the non-emphatic negation of a lexical main verb in a declarative main clause. We exemplify it with English. ()

Mary does not love him

[standard negation]

Negation in all other functions is referred to collectively as ‘non-standard negation’, illustrated in (). ()

a. b. c. d. e. f. g. h. i.

Mary does not love him at all Mary does not live here yet I urge you not to talk to him Doesn’t Mary love John? Don’t listen to him Fred is not a teacher There are no blue tigers There are no blue tigers in France Nobody believes him

[emphatic negation] [phasal negation] [subordinate negation] [interrogative negation] [imperative negation] [ascriptive negation] [existential negation] [locational negation] [negation of indefinites]

¹ Typological survey chapters similar to this one are Dahl () and Miestamo (). Dahl () allots more space than either Miestamo () or us on expression types and word order. Miestamo () is recommended for a brief historiography of the typology of negation, an elaborate discussion of (a)symmetry and brief discussions on subordinate, interrogative, derivational and prosentential negation. Our survey stands out for its focus on multiple exponence, negative indefiniteness, and on interlacing synchrony with diachrony.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



       j. k. l. m.

No! Mary disagrees with me He was without money Don’t be surprised if it doesn’t rain (meaning ‘Don’t be surprised if it rains’,

[prosentential negation] [derivational negation] [privative negation] [expletive negation] Horn a: )

The labels and the illustrations should be sufficiently clear. Note that it is rather common for a language to use the same formal means for several of these uses. In English, standard negation brings into play the negator not (or n’t) and when there is no auxiliary or copula or when the verb is not the lexical verb have, we generally also get the ‘periphrastic’ auxiliary do. And the same is true for the negation in (a–b, d–e, m). But often a language uses a different strategy, like in English (c, i–l). Another worthwhile point is that (k), (l), and (m) are not actually ‘negative sentences’. In (m) doesn’t arguably has no meaning. (k) and (l) are not negative sentences, because they are positive, though the predicates they ascribe to the subject are negative. Klima () is the classical discussion of tests for showing what is a negative sentence and what not. One such test for showing that (k) is positive consists of continuing it with a positive and so does instead of the negative and neither does, appropriate for (). A similar test works for (l). ()

a. Mary doesn’t love John and neither does Suzy b. Mary disagrees with me and so does Suzy

Section . deals with aspects of standard negation. In section . we focus on the three beststudied types of non-standard negation, viz. prohibitive negation, existential negation, and the negation of indefinites, though there will be side remarks on other types too. The issues that will come up also appear in the other chapters, and some overlap is unavoidable.² Of course, the approach in this chapter is typological, with the objective to lay bare some of the variation found in the world’s spoken³ languages, synchronic as well as diachronic. To some extent we also make statements on what is more or less frequent in the world’s languages, on what is decidedly rare or dominant—without statistical sophistication, however—and we provide typological explanations (often partially diachronic) for these observations or hypotheses. Ideally, statements on what is universal, frequent, or rare are sample based, that is they should be based on an attempt to provide some kind of representative data. However, some of the observations and hypotheses are based on large data sets where the collection has not followed any sampling strategy.⁴ These data sets are sometimes called ‘convenience samples’, but we will consistently refer to them as ‘data sets’. They are a less trustworthy source for finding out about linguistic diversity. However, they tend to be much bigger than the samples, and for this reason the general tendencies as to what is frequent or rare are likely to be visible there too. Given the

² Phasal negation may well stay under the radar in this book. See Van Baar () and Kramer (forthcoming). ³ On signed languages, see Zeshan (a), Zeshan (), and Oomen and Pfau (). ⁴ On sampling for typology, see Miestamo, Bakker, and Arppe () and the references therein.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

   



increased importance of the subfield of areal typology, we will also venture to make areal statements.

.. S 

.................................................................................................................................. Negation is a superficially simple semantic operation. The negative sentence has exactly the same meaning as the positive one, except for the effect of negation. Thus a positive declarative says that some state of affairs holds and its negative counterpart presents the very same state of affairs but says that it does not hold. One would thus expect that this ‘simple meaning’ is expressed with a simple strategy. Intuitively, simplicity has two sides to it.⁵ First, a simple meaning would require just one marker. Second, the proposition within the scope of the negation would be expressed in the same way as in the positive sentence. We discuss the first issue in section .. and the second one in section ... We then turn to the kinds of markers languages use for negation (section ..) and to their placement in the sentence (section ..).

... Single vs. multiple exponence In the world’s languages standard negation is indeed usually expressed with one negator only, like in (), and unlike in (). ()

French Marie ne le voit pas Marie  him sees  ‘Marie does not see him’

This has been confirmed in several studies. For a worldwide sample of  languages Van Alsenoy (: ) found that  (%)⁶ have a single exponence strategy only. In two large data sets, we find similar results. First, in a database of , Austronesian and SinoTibetan languages and the languages of New Guinea, Australia, and the Americas, Vossen () found single exponence in , (%) languages.⁷ Second, Dryer (a) has a worldwide data set of , languages of which , (%) may well use a single negator only.⁸ This still leaves a good many languages that may or must have more than one negator. The most common type has two negators, either optionally or obligatorily. Thus in

⁵ Despite this appeal to intuition, simplicity in grammar is not an easy matter. On simplicity and complexity in typology, see Miestamo, Sinnemäki, and Karlsson (). ⁶ We simplify percentages to full digit numbers. ⁷ Here and elsewhere we are grateful to Frens Vossen for having calculated figures on the database underlying Vossen (). ⁸ The hedge with may well is due to the fact that Dryer’s numbers are not restricted to standard negation. He would seem to include interrogative, ascriptive, existential, and locational negation and perhaps also emphatic negation.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



      

Vossen’s dataset of  languages that may not or cannot suffice with one negator, all but eight do not allow more than two and in Dryer (a) the numbers are very similar. The French example in () illustrates obligatory doubling for written standard high register French, but it also illustrates optional negation, for in all other types of French ne can be dropped. It is clear from Vossen () that triple and quadruple negation are always optional and rare, quadruple rarer than triple, and we know of only one case of quintuple negation. ()

Kanyok (kany, Devos and van der Auwera : )9 ka-tù-tùm-in-oh’ bènd --send--  ‘We have not sent’

()

Lewo (lewo, Early a: ; b: ) pe-re a-pim re poli - -come.   ‘They didn’t come’

()

Bantawa (bant, Doornenbal : ) i-ciŋ-nin set-nin-ci-n -hang- kill--10- ‘He does not kill himself by hanging’

()

Bantawa (bant, Doornenbal : ) i-ciŋ-nin set-nin-Ø-nin-ci-n -hang- kill----- ‘He is not killing himself by hanging’

Note that the Bantawa cases of quadruple and quintuple negation involve two verbs building what the specialist literature (Doornenbal ) calls a “compound verb.” Bantawa is a central Kiranti language. Since central and eastern Kiranti languages (kira) all have double or triple negation (van der Auwera and Vossen ) and since both verbs of the verbal compound attract negative marking, we can get up to five markers. Why do languages bother about having two or more negators? The answer given by typologists refers to what is commonly called a ‘Jespersen’s Cycle’ (or ‘Jespersen Cycle’) and less commonly also ‘Negative Cycle’. The terms with the proper name are due to Dahl () and the reference is to the opening lines of Jespersen (). It is important to realize, however, that (i) Jespersen was not the first to propose the explanation named after him, ⁹ In the glosses the Leipzig glossing rules () are used, but we stay as close as possible to the original description. We keep the orthography of the example as found in the source. For less familiar languages and language families we supply ‘glottocodes’ () for unique identification, since many languages and language families have different names. When there is no source indication, the grammaticality judgments are ours. ¹⁰ See Doornenbal (: ) on why we find ‘’ to be a good gloss.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

   



(ii) there is now a multiplicity of Jespersen Cycles, unimagined by Jespersen (), which are sufficiently different from one another to drop the singular in ‘Jespersen Cycle’ and to opt for a plural ‘Jespersen Cycles’ (van der Auwera ; van der Auwera, Krasnoukhova, and Vossen (forthcoming)), and (iii) not every Jespersen Cycle yields a multiple negation. The textbook illustration of a Jespersen Cycle shows French ne . . . pas. French inherited a preverbal ne from Latin, and even in Old French ne was often accompanied by a minimizer, an expression that refers to a small entity or quantity, like pas ‘step’, point ‘point’, or miette ‘crumb’. The effect of adding a minimizer was pragmatic, probably emphatic: a state of affairs did not just not obtain, it did not even obtain in a minimal form. The minimizing and emphatic effect then bleached, and one of the original minimizers, viz. pas, became a near-obligatory component of negation. Thus the negation became double. ()

ne > ne . . . (pas) > ne . . . pas

The three stage model in () is not the whole story, however. Even for the high registers of written French, which has ne . . . pas, there is a controversy as to whether ne is still negative (see already Jespersen : ), and, as mentioned, for some registers, ne is not necessary anymore. Thus we can add a fourth stage and a fifth stage. ()

ne > ne . . . (pas) > ne . . . pas > (ne) . . . pas > pas

One hundred years after Jespersen () there is a fair amount of agreement that the French scenario is just one highly specific instantiation of the general phenomenon. It is true that linguists have considered one or more properties of the Jespersen Cycle à la française to be definitional, but it is best to regard them as optional (see also van der Auwera, Krasnoukhova, and Vossen (forthcoming)). Here are the important features of the French ne . . . pas cycle. ()

a. b. c. d. e. f. g. h. i.

in the middle of the cycle there are two negators, viz. ne and pas the new negator is originally something other than a negator the old negator disappears the beginning and the end of the cycle have one negator, but it is a different one the negators are syntactic elements the negators are on different sides of the verb the new negator is more to the right than the old one a Jespersen cycle is a language-internal change the reason for this process is the phonetic reduction of the first negator

We now discuss each of these properties. ad (a): Is doubling necessary? To answer this question it is useful to have a look at Greek. Greek replaced an old negator ou with a new one, the latter consisting of the old negator ou and the phrase de hen, meaning ‘even one’, but this phrase never became a negator. Instead it merged with the old one, giving ouden and it is this merger that became the new negator, later simplified to den. Willmott () and Chatzopoulou (, )

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



      

both analyzed the diachrony of Greek negation. For Wilmott (), who considers doubling to be essential, Greek does not have a Jespersen Cycle, though she notes the similarities between a Jespersen Cycle and what happened in Greek. Chatzopoulou (, ) proposes a very similar analysis, but she does not consider doubling to be crucial—and neither does Schwegler (, )—and this way she does propose a Jespersen Cycle for Greek. Both analyses are acceptable, but the better one is the Chatzopoulou–Schwegler approach, as argued by van der Auwera, Krasnoukhova, and Vossen (forthcoming). ad (b): Does the new negator have to come from something non-negative, like in French, which has pas going back to the noun ‘step’, or like in Avava, with -mu deriving from ‘first’? ()

Avava (katb, Crowley : ) na-sa-robit-mu .--hear- ‘I didn’t hear (it)’

Here the answer is uncontroversial. Nearly as classical a Jespersen Cycle as that of French is that of English, with the new not negator joining the older ne negator for emphasis, then bleaching and replacing it; this not derives from a pronoun meaning ‘nothing’. () is a Middle English example showing a ne . . . not construction in which the not word is still emphatic. () Middle English (Ingham : ) Ne ic ne cume to heom nawiht and I  come to them  ‘I do not come to them at all’ () is an example from Biak. Different from English, in Biak it is the first negator that provides emphasis and it was a clausal negator already, but in a different language, viz. in the local Malay or Indonesian (Van den Heuvel : ). ()

Biak (biak, Van den Heuvel : ) indya bukan ko-kain ko-fafyár biasa va so  .-sit .-tell usual  ‘So we are not (just) sitting and telling her (but have a serious meeting).’

ad (c): The old negator need not disappear. In some Dutch dialects it became expletive and survived as a subordination marker (van der Auwera : , also for references, and Van de Velde and Norde : – for an account in terms of exaptation). () Katwijk Dutch (van der Auwera : ) Toen we bij de poort en kwamme . . . when we at the gate  came ‘When we arrived at the gate . . . ’

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

   



With tripling, the old negator also does not disappear, for double exponence does not have to go back to single exponence, but can instead develop into triple exponence, as in Kanyok (). ad (d): The new negator need not be different from the old one. Numerous are Jespersen Cycle analyses of a second negator being a copy of the first one. Brabantic Belgian Dutch usually negates with a single postverbal nie (i.e. postverbal relative to the finite verb—see ad f below) but some speakers can add a clause-final copy. () Brabantic Belgian Dutch Ik heb hem nie gezien nie I have him  seen  ‘I haven’t seen him’ () shows tripling in Mandan: it is emphatic due to the clause initial doubling of the negative prefix wa:-. ()

Mandan (mand, Mixco : ; Vossen : –) wa:-wa:-i=ra-sæk-ri̜x-oʔš --=-do-- ‘You didn’t work’

ad (e): Negators involved in a Jespersen Cycle may, of course, be syntactic elements, like in French (), English (), Biak () or Dutch (), but they may be morphological as well, just like negators that stay out of the cycle. The morphological vs. syntactic status of negators will be discussed in section ... So far, we have seen morphological negators in Bantawa () and () and in Mandan (), where all negators are morphological, but also in Kanyok () and Lewo (), which combine morphological and syntactic negators.¹¹ ad (f ): Jespersenian doubling is overwhelmingly verb-embracing, as illustrated by French () and English (), but in Oneida both negators precede the verb. ()

Oneida (onei, Abbott : ) yah teɁ-ho-nohale-Ɂ  --wash- ‘He did not wash it’

Brabantic Belgian Dutch () could be taken as an example of two postverbal negators (see De Swart :  for this claim for similar structures in Afrikaans): at least both nie negators follow the finite verb and, as we will argue in section .., it is the finite verb that is relevant for Jespersenian doubling. However, in () both negators do embrace the other verb, the non-finite one. In any case, it is clear that a claim that doubling negators should embrace a finite verb in the French way is problematic. An embrace account is also problematic for multiple exponence: in Lewo (), for instance, there is an embrace, but it ¹¹ In Lewo (6), moreover, the syntactic pe-re negator morphologically breaks down in the two negators pe and re.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



      

is heavier on the right side with two separate negators contrasting with a univerbation on the left. ad (g): A Jespersen Cycle need not progress from left to right. From the data in Vossen (; see also van der Auwera and Vossen ) it seems that the left to right direction is the more frequent one, but we know of numerous cases of a Jespersen Cycle ‘in reverse’. For example, in the Awju-Ok languages (awju) there is family-internal evidence for regarding the rightmost negator of double exponence structures to be the oldest one and the leftmost negator the newer one (Vossen : –). Thus the way negation marked in Mandobo is hypothesized to reflect an earlier stage compared to the one in Tsaukombo. ()

Mandobo (mand, Wester : ; Drabbe : ) ro otogop-gen do, kerewatop nda form see-.  face  ‘He looks and they do not have faces’

()

Tsaukambo (tsak, de Vries : ) bɔ-mɛlixɛ-nda -go.home- ‘I do not go home’

Biak () offers another illustration: the clausal final negator va is the older one, it is borrowed from surrounding Papuan languages (Reesink a: ), while bukan is a more recent borrowing from Malay or Indonesian (Van den Heuvel : ). These borrowing facts lead us directly to feature (h). ad (h): In French, the Jespersen Cycle is a languageinternal phenomenon in the sense that there is no need for invoking influence from another language. But this is not the case for Biak. The va negator is borrowed from Papuan and its clause-final position is the typical one for the Papuan languages, not for Austronesian languages. The bukan negator is borrowed together with its position, too, in this case the Austronesian default preverbal position. For another example of the relevance of language contact, we can go to South Vietnam, where the Austronesian Chamic languages (cham) have been spoken in close contact with the Austroasiatic Bahnaric ones (bahn) for two millennia (Thurgood ): in these families we can have the same negators, though it is not obvious from which family they spread, and we also see doubling. This is exemplified with Chamic Jarai and Bahnaric Rengao. ()

Jarai (jara, Lee : ) kão bu homão prǎk ôh I  have money  ‘I don’t have any money’

() Rengao (reng, Gregerson : ) aw bIq loq Oh I  know  ‘I don’t know’

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

   



Van der Auwera and Vossen () make the case that the doubling pattern itself was calqued (from Chamic to Bahnaric). ad (i): For Jespersen () the reason for the appearance of the new negator is the phonetic weakness of the old one. It is not disputed that this may be at work in some languages, but at least for French, a second and better analysis was already offered by Meillet (). For Meillet, ne itself was fine for ordinary negation, but pas originally realized an emphatic negation, something like ‘not at all’ (from ‘not a step’), which then bleached and became neutral too (he took the cycle, which he called a “spiral,” to illustrate the general process for which he introduced the term “grammaticalization”—in that same chapter). Also, if doubling can be calqued, the reason may just be the prestige of the donor language. Yet a fourth explanation relates to constructional asymmetry (see section ..). And, finally, there is no reason to assume that there should only be one motivation. Areally, the Jespersen Cycle is attested over the entire globe, but it is sparse in Eurasia, except for its Standard Average European corner. The claim that the Jespersen Cycle would be one of the features characterizing Standard Average European goes back to Bernini and Ramat () and we indeed see the Jespersen Cycle, for instance, in French and in Italian dialects (Vossen : –), but not in European Portuguese or Romanian and not in Slavic or Indo-Aryan (van der Auwera : –). In other areas and phyla there are concentration zones as well. For Austronesian, for instance, there are three clusters: the Chamic cluster, already mentioned, in Vietnam and Cambodia, New Guinea and especially Vanuatu, and for Sino-Tibetan the ‘hotbed’ is Nepal with its central and eastern Kiranti languages (Vossen : –). A further comment concerns the possibility of zero exponence. Paradoxically, the simplest method of ‘marking’ negation is not to mark it all—and ‘mark’ the absence of negation instead. We will come to this in section .. and accept the possibility of zero marking. But zero marking, that is the absence of a marker, constitutes marking too, if it contrasts with a paradigmatic alternative, which is then affirmative rather than negative. A related question is whether every language has standard negators. The answer in the literature is implicitly positive. Note there is at least one grammar that gives an implicit negative answer, viz. the grammar of Ese Ejja (esee), as described by Vuillermet (: –): it would have a phasal ‘not yet’ and a ‘never’ but no standard negator. Also, we want to stress that one can express negation in a declarative clause with a lexical main clause verb without a standard negator. This is what English does with a negative indefinite, as illustrated in (i). We will come to this in section ... It is important to realize that saying that a language expresses negation with a single exponent means that there can only be one exponent of a standard negator in every sentence, not that there can’t be different standard negators depending on different kinds of sentences. By the same token, languages may also have alternative multiple negators. In Vossen’s () dataset of , languages with single exponence about  (%) have this kind of alternation, but it is often difficult to judge what counts as an alternative standard negator (instead of a non-standard one). Though there is no systemic study to date, it is clear that the alternation typically depends on tense, aspect, or mood. Thus Fe’fe’ has three standard negators, one double and two single negators, and they alternate depending on tense and aspect: sı .̍ . . bά for the non-past and habitual, sì by itself

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



      

for the non-hodiernal past, and kàɁ for the hodiernal past and the perfective present (Ngangoum : , –). ()

Fe’fe’ (fefe, Ngangoum : , , ) a. Siani sı ̍ ká γέ nte ̍e ̍ bα̍ Siani   go market  ‘Siani will not go to the market’

[non-past]

b. Siani lὲ sı̀ γε̍ nte ̍e ̍ Siani .  go market ‘Siani did not go the market’

[non-hodiernal past]

c. Siani kὰɁ fhú ŋgέ ntee̍ ̍ Siani  . go market ‘Siani has not gone to the market’

[hodiernal past]

An illustration for mood alternation is offered by Nanti. It uses te(ra) for realis negation and ha(ra) for irrealis negation (Michael ). ()

Nanti (nat, Michael : , ) a. Tera i=N-poroh-e . .=-clear.land-. ‘He is not clearing land’ b. Ha=me pi=tsot-se-na-i=ro .= =slurp.up--.-.=. ‘You shouldn’t slurp it up’

A final point is that the doubling that is typical, though not necessary, for a Jespersen Cycle is different from the doubling in what has been called ‘negative concord’ constructions, such as (). ()

We don’t need no education

This phenomenon as well as the link with Jespersen doubling will be discussed in section ...

... Symmetry and asymmetry If one compares the German affirmative and its negative counterpart, one can see that the two assertions differ only in one respect. The negative sentence adds the negator nicht, everything else is the same and this simple strategy is not only used in the simple present, as in (), but in all tense-aspect-mood-voice combinations. ()

a. Marie liebt ihn b. Marie liebt ihn nicht

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

   



Miestamo () calls this a negation strategy that is “symmetric” both “constructionally,” because the two constructions differ only as to whether nicht is present or absent, and “paradigmatically,” because it is found in the entire verbal paradigm. This is a simple system, but languages can be more complex, both constructionally and paradigmatically. Constructional ‘asymmetry’ can be illustrated with Carib. ()

Galibi Carib (gali, Mosonyi, Mosnoyi, and Medina Tamanaico : , ; Miestamo : ) a. m-oonaae -cultivate. ‘You cultivate’ b. oona-ja maana cultivate- . ‘You don’t cultivate’

(b) differs from (a) not only in having a negator ja, but also in that the lexical verb now appears in a non-finite form accompanied by a copula. This is not too different from English: in the translation of (b) the form cultivate is an infinitive and there is a periphrastic do verb, absent in the positive sentence.¹² An illustration of paradigmatic asymmetry is given with Kresh. In the affirmative the language distinguishes between a perfective, an imperfective, and a perfect, but in the negative there is only a perfective, which negates not just the positive perfective, but the positive imperfective and perfect as well. ()

Kresh (kres, Brown : ; Miestamo : –) a. kôkó ãnjã mömõ Koko he.go home ‘Koko went home’ b. kôkó ãnjã mömö ĩshí Koko he.go home he.have ‘Koko has gone home’ c. kôkó ă yãnjã mömõ Koko at act.of.going home ‘Koko is going home’ d. kôkó ãnjã mömõ ‘dĩ Koko he.go home  ‘Koko didn’t go / hasn’t gone / isn’t going home’

Both asymmetry types come in subtypes and have been argued to need a variety of language-particular, diachronic, and/or functional explanations. Thus the appearance of the non-finite verb with a copula in (b) has been claimed to be a reflection of the stative character of negation (Givón : ; Hagège : –; Miestamo : –, –): ¹² Miestamo (: –) does not analyze English this way, however, because there is also an emphatic use of do, as in You do cultivate, which (b) is constructionally symmetric with.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



      

when one denies a state, one gets a state, but when one denies an event, one typically gets a state too. See the examples in (), taken from Miestamo (: ). ()

a. b. c. d.

Chris knows the song Chris doesn’t know the song Chris drank the coffee Chris didn’t drink the coffee

[state] [state] [event] [state]

The neutralization illustrated in Kresh may be a reflection of the fact that negative sentences typically occur in contexts in which a corresponding positive sentence is present or assumed. Thus the aspect neutral (d) will typically occur in a context which contains or assumes one of the three aspect specific positive sentences (a–c), and thus there is less of a need to repeat this information in the negative. Asymmetry is by no means rare in the languages of the world. In his -language sample Miestamo (: ) finds the constructional type in % of his languages and the paradigmatic type in %. The dominance of constructional asymmetry over paradigmatic asymmetry holds for the whole world, except for the larger Pacific area (with Southeast Asia, Oceania, Australia, and New Guinea), in which the two types are equally common (Miestamo : ; see here for more areal observations). A final point takes us back to the Jespersen Cycle. In constructional asymmetry, the negative sentence is distinguished from the positive one by the negative marker as well as by one or more other markers, not themselves originally expressing negation. But these markers may be reanalyzed, not unlike how French pas ‘step’ was reanalyzed. This is illustrated with Arizona Tewa. () Arizona Tewa (ariz, Kroskrity : ) we-mán-mun-dí Sen kwiyó man woman ->.-see- ‘The man did not see the woman’ Originally, restating Kroskrity () using Miestamo’s asymmetry framework, () was a constructionally asymmetric sentence, with a negator we forcing a subordinator dí on the lexical verb. This subordinator was later reinterpreted as a negator in its own right.

... Types of expression In the preceding sections, the examples have already shown that there are different ways of marking negation. Mandobo (), for instance, has a syntactically free negator and Tsaukambo () a bound one. The former is more frequent than the latter and globally so, with concentration zones for the bound ones in central Africa, north-east India and Nepal, and northern South America, an estimate based on Dryer (a) and Vossen ().¹³ On a total of  languages with a single negative marker which he feels confident

¹³ The number does not tell us anything about the in-between category of clitics, for they are included in the figures for free negators in Dryer (a), our main source.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

   



about classifying,¹⁴ Dryer (a) has  (%) languages with a free exponent and  (%) with a bound one. The syntactically free negators are usually particles ( languages (%)) but negators can also be auxiliaries, which is typical for northern Eurasia (Dryer a), as illustrated with Evenki (). () Evenki (even, Nedyalkov : ) Eveŋki-l e-ŋki-tin utele gule-l-ve ō-ra Evenk- -.- formerly house-- make- ‘The Evenks did not build houses previously’ An example of a language that expresses negation by tone is Eastern Oromo, but tonal expression is rare and even rarer if it is the sole exponent of negation. Out of a total of , languages Dryer (b) found tone on the verb only in seven languages, all of them African—the tone is marked with the accent symbol on jír. ()

Eastern Oromo (east, Owens : , ; Miestamo : , ) a. inníi deem-úu jira he go- exist. ‘He is going’ b. inníi deem-úu-ti he go-- ‘He is not going’

n-jír-u -exist-

Dryer (b) also mentions stem alternation (attested in Berber, Lafkioui and Brugnatelli forthcoming) and infixation (as in Bantawa () and ()) and they are very rare. What he does not mention is exponence by nothing or, better, by a zero morpheme. There is no question that it exists, be it on a very limited scale. Thus in Havyaka Kannada—and elsewhere in South and Central Dravidian, the negation of a non-past subjunctive is expressed by the absence of a filler of the tense slot, a phenomenon that has attracted scholarly attention since at least Master (). () Havyaka Kannada (nucl, van der Auwera and Bhatt : , , ) a. ma:’sTrakko kate o:du-g-u teachers story read-.- ‘The teachers may be reading a story’ b. ma:’sTrakko kate o:d-Ø-avu teachers story read-.- ‘The teachers may not be reading a story’ It has also been proposed as a more general strategy. After all, if the difference between an affirmative and a negative main clause declarative can be marked by something in the negative, why couldn’t it be marked by something in the affirmative?

¹⁴ For an additional 73 languages it is unclear whether the negator is free or bound and a further 21 languages have both a free and a bound negator.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



      

()

Form Meaning ‘Mary loves him’ ‘Mary doesn’t love him’

Difference marked in the negative affirmative Mary love him Mary AFF love him Mary NEG love him Mary love him

Affirmative marking, in the sense described above, however, either does not exist or is exceedingly rare. The case discussed most widely is the one illustrated with Karitiána (). This language does have a negator (padni) but it is frequently omitted (Storto ; Everett : –). A non-occurrence of mood markers in negative clauses in Karitiána coupled with the elision of the negator results in sentences like (b) (see Miestamo  for a typological discussion of negatives without negators). () Karitiána (kari, Landin : ) a. Y ta-oty-j ỹn I -bathe- I ‘I will bathe’ b. Y oty ỹn I bathe I ‘I will not bathe’

... Position It has been suggested as early as by Jespersen (: ) that [T]here is a natural tendency, also for the sake of clearness, to place the negative first, or at any rate as soon as possible, very often immediately before the particular word to be negatived [sic] (generally the verb).

This statement contains a few hedges (at any rate, as soon as possible, very often, generally the verb) and some unclarity. That there are these hedges makes sense, because the paragraph immediately follows Jespersen’s observation that French developed a postverbal negative pas. The unclarity concerns the notion of verb: in case a sentence has both an auxiliary and a lexical verb, it is not made explicit what the reference point is for calling a negator preverbal or postverbal. The context, however, undoes the unclarity: Jespersen must have meant the finite verb: it is because pas follows a finite verb, whether a lexical verb or an auxiliary, that Jespersen considers pas postverbal. Another problem is that to propose that there would be a natural tendency for an early placement of the negator Jespersen had few data, and neither had Horn (), who canonized Jespersen’s conjecture with the term ‘Neg-First’ (principle). But meanwhile the conjecture/principle did get cross-linguistic support. Thus in Vossen’s () dataset of , languages with single exponence, something close to  languages (%) have the negator in preverbal position.¹⁵ Some support

¹⁵ The hedge with something close to is due to the fact that Vossen (: ) includes negative verbs/ auxiliaries.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

   



can also be found in Dryer (b), with the proviso that ‘preverbal’ is here defined relative to the lexical verb: of the , languages in Dryer’s dataset that mark their single negation consistently either before or after the verb,  (%) have the negator in preverbal position.¹⁶ Nevertheless, there are families like Altaic that have a lot of postverbal negation, as well as areas, including New Guinea (Reesink b; Klamer, Reesink, and van Staden ; Vossen : , ), the ‘Macro Sudan Belt’ (Güldemann ), and South America (Muysken et al. : –; Vossen : ). There are also correlations between the position of negation and other word order properties of the languages as well as with the presence of a Jespersen Cycle. Thus, for example, Dryer (c) reports the following tendencies: in SVO languages the dominant positions of negation are SNegVO and SVONeg; in SOV languages SONegV and SOVNeg are dominant; in verb-initial languages NegVSO and NegVOS patterns prevail, and the ONegVS pattern is most common among object-initial languages (cf. Dryer c for discussion and some tentative explanations, as well as Dahl : –; Dahl : –; Dryer : –; and Dryer b).

.. N- 

.................................................................................................................................. Making a distinction between standard and non-standard negation is to some extent making a distinction between the type of context the negator occurs in. For all types of non-standard negation except for expletive negation, the meaning is still the same, although it would have to be discussed in more general terms than, for example, saying that negation reverses the truth-value. The negator in () obviously does not change any truth-value, for imperatives do not have truth-values. ()

Do not believe him

In non-standard negation the properties of the negators may be identical or similar to those found in standard negation. Thus the not in () is very similar to that in (). ()

I do not believe him

In both contexts the form is the same, not allows the short form n’t in both contexts, and it appears between the do auxiliary and the lexical verb, the latter appearing in the infinitive. But note that the two not negators are still different. In a declarative the copula does not allow do, but in prohibitives do is obligatory. ()

a. You are not happy b. Don’t be happy

Also, when the prohibitive has a subject, this only allows the short form of the negator, which furthermore allows a univerbation with the subject pronoun.

¹⁶ Dahl (: ) judges an earlier count by Dryer, viz. Dryer (), and is mildly optimistic that some version of the Neg-First principle does indeed hold.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

 ()

       a. b. c. d. e.

You do not believe him You don’t believe him *Do not you believe him Don’t you believe him Dontcha believe him

In the following sections we will discuss some of the non-standard negation uses with a focus on how they affect the properties on their negators. We will start with the type illustrated in the above, that is the prohibitive.

... Prohibitive negation In English the differences between standard and prohibitive negators are subtle. In most languages, however, the differences are obvious. Van der Auwera and Lejeune () offered a four way typology, based on whether the verb of the prohibitive is the same as the verb of the imperative and on whether the negator is the same as the one in standard negation. English imperatives and prohibitives have identical verb forms and, despite the subtle differences, both prohibitive and standard negation use identical negators. In a data set of  languages,  (%) languages are like English, and, they claim, like Standard Average European. So in most languages either the verbs in imperatives and prohibitives are different or the negators in standard and prohibitive negation are different, or both are different. In their data set the most common type has a special negator only ( languages or %), as illustrated in (). ()

Vietnamese (viet, Thompson : ; Victoria Rosén, p.c.) a. Uông ruou! drink alcohol ‘Drink alcohol!’ or ‘I/you/he/etc. is/are drinking alcohol’ b. Chó uônr ryou!  drink alcohol ‘Do not drink alcohol!’ c. Không uông ruou.  drink alcohol ‘I/you/he/etc, is/are not drinking alcohol’

Then comes the type with both special negators and verb forms ( languages or %). ()

Zulu (zulu, Poulos and Bosch : ; Khosi Mnyakeni, p.c.) a. Shay-a inja! hit-. dog ‘Hit the dog!’ b. Mus-a uku-shay-a inja - -hit- dog ‘Do not hit the dog!’

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

   



c. U-ya-shay-a inja -.-hit- dog ‘You hit the dog’ d. A-wu-shay-i inja -.-hit-.. dog ‘You do not hit the dog’ The last type has special forms only ( languages or %). () Spanish a. Pedro canta Pedro sing... ‘Pedro sings’ b. Pedro no canta Pedro  sing... ‘Pedro does not sing’ c. Canta! sing.. ‘Sing!’ d. No cantes!  sing... ‘Don’t sing!’ A variety of partial explanations for this distribution have been offered. Thus the Spanish subjunctive in (d) has been claimed to be more indirect than the imperative and thus, to vary on what Horn (: ) wrote about negative modality, “ ‘cushions the iron fist’ of prohibition ‘in the velvet glove’ of the description of what is merely wished for” (van der Auwera : ). The greater need for indirectness and securing it through renewal would also explain why prohibitives exhibit more variation than imperatives (Van Olmen : ; Devos and Van Olmen ). For reasons of space, several issues remain untouched. Thus we do not discuss first and third person constructions, such as (), though they are sometimes treated as imperatives and prohibitives, too.¹⁷ Furthermore, we do not discuss tense aspect issues, the question of

¹⁷ There is, for instance, the question whether past ‘prohibitives’ such as Estonian (a) are truly prohibitives. See Van Olmen () for a typologically sustained negative answer. (a)

Estonian (Van Olmen : ) No är-nud too-nud siis, keegi ei  -. bring-. then nobody  käski-nud ju command-.  ‘Well, you/he/ . . . should not have brought it, no one told you/him/ . . . !’

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



      

whether a Jespersen Cycle operates in the same way for prohibitives or what the result of a Miestamo (a)symmetry study would be like (Miestamo and van der Auwera ). ()

a. Let’s talk! b. Let’s not talk! c. Don’t let’s talk!

... Existential negation Languages may express the negation of existence with a marker that is different from the standard negator. In Veselinova’s sample of ninety-five languages (: ) this is the case in forty-two languages (%). Turkish is a case in point. In (a) the -me affix is the standard negator, and in (d) there is an existential verb-like negator yok-. ()

Turkish (Van Schaaik : –, –; Veselinova : –) a. Gel-ecek come- ‘She will come’ b. Gel-me-yecek come-- ‘She will not come’ c. Su var-dı water exist- ‘There was water’ d. Su yok-tu water - ‘There was no water’

In another twenty-one languages, the negator is the same but, depending on the function, it has different morphological or syntactic properties. Kannada illustrates how the morphosyntax can distinguish the two uses. In standard negation illa is an affix, but in existential negation it is a free form. ()

Kannada (nucl, Sridhar : , ; Veselinova : ) a. Anil ka:le:jige ho:gu-vu-illa Anil college. go-.- ‘Anil won’t/doesn’t go to college’ (< ‘There is no Anil going to college’) b. Khaja:neyalli haNa illa treasury. money  ‘There is no money in the treasury’

On the basis of this ninety-five-language sample one might thus hypothesize that roughly two-thirds of the world’s languages have negators used for existential negation that differ from standard negators. That does not mean that these negators are uniquely used for existential negation. Veselinova (: ) identifies as many as twenty-one uses, different

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

   



from standard negation, that may be shared by the negators that are used for existential negation. For instance, it is common for negators that deny existence to also deny possession ( of  languages), less so location ( of  languages) and sometimes they also have the prosentential ‘No!’ use ( of  languages). Veselinova (: ) also shows that the negators used for existential negation but not for standard negation are widely spread across the globe, more so than grammaticalized expressions for existence, and that there is no evidence that the constructions for existence and the negation of existence are related in any strong way. These negative existentials have two types of origin: they either result from a univerbation of a (standard) negator and an existence marker or they result from a lexical item with a negative meaning such as ‘lack’ or ‘empty’ (Veselinova : –). What has been studied most is what is called the ‘Negative Existential Cycle’, first proposed by Croft (), then extensively studied by Veselinova (, , ,  and Veselinova and Hamari, forthcoming). The Cycle is summarized in (). ()

one negator is used for both standard and existential negation → one negator is used for standard negation and another one expresses existential negation → one negator is used for both standard and existential negation, but it is the one that was previously only used as the existential negator and so for standard negation it is a ‘new’ one

Just like the Jespersen Cycle, the Negative Existential Cycle has two intermediate stages, for a language can be on the move between the three stages of (). In terms of this typology, Turkish is thus in the second stage and Kannada has been analyzed as either in stage  (for literary Kannada) or in between stage  and  (for spoken Kannada) (Veselinova : –, –). Veselinova’s ninety-five-language sample suggests that some % of the languages complete the cycle, % have not started it, and the rest are in between (Veselinova : ). Veselinova (: ) also shows that the presence of the Negative Existential Cycle is strongly family based: where stage  is the worldwide winner, it is ‘as good as absent’ in Turkic, Dravidian, and Polynesian. Unlike for the Jespersen Cycle, there is no evidence that existential negators allow multiple exponence, though standard negators deriving from existential negators may be involved in multiple exponence, that is for standard negation (van der Auwera, Krasnoukhova, and Vossen, forthcoming). As for standard negators, languages may have more than one existential negator. Currently under discussion is whether privative and ascriptive negation also enter into Cycles like the Negative Existential Cycle. At least for Arawakan languages (araw) and perhaps also in the Takanan language Tacana (taca) it has been claimed that standard negators derive from privative negators (see Michael  for Arawakan and Guillaume  for Tacana). The similarity between privatives and negative existentials is obvious: when a state of affairs is without something then this something does not exist in the state of affairs. In ten of the ninety-five languages of Veselinova’s sample, privatives and negative existentials are the same (Veselinova : ). One could thus consider a Privative Cycle to be a subtype of the Negative Existential Cycle.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



      

There could well be a ‘Negative Ascriptive Cycle’ too. If () sketches the reanalysis of the Negative Existential cycle, then () would be the reanalysis of a hypothesized ‘Negative Ascriptive Cycle’. ()

[there is no] Anil going to college

->

Anil does [not] go to college

()

Anil is a [not-goer] to college

->

Anil does [not] go to college

Given that Barnes (: ) considers the negation in () to be “sentence negation,” which must be our “standard negation,” and given her ascriptive paraphrase, the possibility of a ‘Negative Ascriptive Cycle’ (see Krasnoukhova and van der Auwera, under review) seems real enough. () Tuyuca (tuyu, Barnes : ) atí-e-go nĩĩ ́-ã-wõ come--.. be-- ‘She never came’ [Literally: ‘She is (habitually) a not-coming one.’]

... Negation of indefinites From the point of view of simplicity and constructional symmetry, the indefiniteness strategy illustrated in (i), repeated as (a), should be considered less than optimal. (a) is truly a negative clause, it is a main clause with a lexical verb, so this construction would count as standard negation if it wasn’t for the fact that the sentence contains no standard negator. It is constructionally asymmetric as well. The positive counterpart is (b). (a) ‘hides’ the clausal negation in the negative quantifier—we will call it a ‘negative quantification’ pattern—and to that extent it is misleading (Haspelmath : ). (a) does not ascribe a property to a subject referred to as ‘nobody’. From the points of view of simplicity and symmetry, the better strategy is the one illustrated in (c). ()

a. Nobody believes him b. Somebody believes him c. Somebody does not believe him

In English, (c) is not ungrammatical but it has a special, pragmatically marked reading. Then the indefinite outscopes the negation and (c) is about somebody in particular not believing him. However, this is only English. There are up to now two typological and sample-based studies on negative indefiniteness and both show that the strategy illustrated in (c) is the most common strategy.¹⁸ In Kahrel (: ) it is found in % of a forty¹⁸ This strategy need not employ a dedicated pronoun like English someone. It may be a noun meaning ‘person’. What is important is that the strategy used for (50b) combines with a negator to yield the meaning corresponding to (50a). Also, the types mentioned in what follows are the major ones, but they do not exhaust the typology treated in our sources (Kahrel ; Van Alsenoy ).

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

   



language sample; in van der Auwera and Van Alsenoy (: ; : ), based on Van Alsenoy (), it occurs in % of a  language sample. Nasioi illustrates this strategy. ()

Nasioi (nasi, Rausch : ; Kahrel : ) a. Nanin nánu-i someone go-. ‘Someone went’ b. Nanin nánu-ari-i someone go--. ‘No one went’

In comparison, the strategy shown in (a) occurs in only % of the world’s languages in both samples. There is an areal skewing as well, from being the most common (%) in North America to very rare (%) in Africa (van der Auwera and Van Alsenoy : ). There are no figures on whether the worldwide favorite, the strategy shown in (c), has concentration or avoidance zones; Standard Average European, however, seems to be a strong avoidance zone (van der Auwera, Decuypere, and Neuckermans : ). English has more to offer, though. A second and a third negative indefinite strategy are shown in (a) and (b): the former is standard English and the latter is widely found in English vernaculars and Creoles, though not in all (van der Auwera ) and, on a global scale, it is most typical for Eurasia (van der Auwera and Van Alsenoy : ). ()

a. I did not see anybody b. I did not see nobody

Both are simpler than the negative quantification pattern of (a), in the sense that they contain the standard negator, but both are still constructionally asymmetrical. In (a) the standard negator is accompanied by a negative polarity item. In (b) it is accompanied by an item that is negative by itself or at least looks like one—a construction that is commonly called ‘negative concord’, a term once again harking back to Jespersen (but to Jespersen : , not ). Cross-linguistically, the negative polarity pattern appears to be more widespread than negative concord—% (Kahrel : ) or % (van der Auwera and Van Alsenoy : ) vs. % (Kahrel : ) or % (van der Auwera and Van Alsenoy : ).¹⁹ The reason why negative concord fares worse than the negative polarity pattern could be that it involves a kind of negative doubling, not unlike Jespersenian doubling. In Van Alsenoy’s sample Jespersenian doubling is found in % of the world’s languages; the figure for negative concord is similar. In languages like French the similarity between Jespersenian doubling and negative concord is strong. First, just as the French Jespersen Cycle is currently undoing the doubling of ne . . . pas ‘not’ to just pas ‘not’, French negative concord is dissolving in that it is replacing ne . . . personne ‘nobody’ by personne ‘nobody’.

¹⁹ This time the samples show a big divergence. The sample by van der Auwera and Van Alsenoy is more than four times the size of the sample by Kahrel and should therefore be more trustworthy.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



      

() a. Je ne te vois pas > I  you see  ‘I don’t see you’ b. Je ne vois personne > I  see nobody ‘I see nobody’

Je te vois pas I you see  ‘I don’t see you’ Je vois personne I see nobody ‘I see nobody’

For this reason the development from ne . . . personne to personne has even been called a ‘Jespersen Argument Cycle’ (Ladusaw : ). Second, personne ‘nobody’ derives from a noun ‘person’ via a negative polarity use (which is still around²⁰). We find the same with the negator pas. It also comes from a noun, this time a noun meaning ‘step’, which turned negative via a negative polarity use. However, there are differences too. For personne the end point is a negative pronoun, but for pas it is a standard negator (or part of one). For personne the evolution goes from a relatively minor construction to another relatively minor construction, while for Jespersenian doubling a relatively minor construction is heading towards the world’s dominant one. Furthermore, the change for personne is not a cycle or a spiral in the sense that the last stage takes us back to the first stage: pas is a single negator like ne was, but personne ‘nobody’ is not a noun meaning ‘person’.²¹ Cross-linguistically, Jespersenian doubling and negative concord do not co-occur very often. In Van Alsenoy’s sample of  languages only twelve have both (Van Alsenoy : ) and only two are like French, in having negative concord independently of word order (see below), in having negative concord for a set of pronouns (rather than just one), and in forbidding Jespersenian doubling and negative concord to co-occur in one sentence and yield a simple negative sense. (), pragmatically strange though grammatical, does not mean that nobody played in the garden. In Ewe, a language that also has both negative concord and Jespersenian doubling, the counterpart does have the single negation use and the combination of Jespersenian doubling (me . . . o) and negative concord of this double negator with an indefinite (ame aɖeke) actually yields tripling. () Personne n’ dans le jardin a pas joué nobody  has  played in the garden ‘Nobody has not played in the garden’ ²⁰ These marginal uses could be taken to show that personne is still a negative polarity item. Neither Kahrel () nor van der Auwera and Van Alsenoy () do that, but if one did, and not just for French then the higher frequency of the negative polarity indefinites vs. the negative ones would be even more pronounced. ²¹ This is a claim about French personne and we are not saying that negative indefinites never develop further. There can be both semantic and formal further developments. We find a meaning-initiated change in French rien ‘nothing’, when it developed the sense of ‘insignificant thing’ (as in un petit rien ‘a small insignificant thing’) or in Jamaican Creole nobadi ‘nobody’ when it developed a free choice use (as in Nobadi we kil nobadi, dem a-go go a kuot ous ‘Anybody who kills anybody has to go to court’)—van der Auwera and De Lisser ). A form-initiated change can be witnessed in some dialects of Brabantic Belgian Dutch, when the negative pronoun niemand ‘nobody’ in a negative concord construction (as in Ik heb niemand niet gezien ‘I have not seen anybody’) dropped the initial n- and became the positive iemand ‘somebody’ (Ik heb iemand niet gezien ‘I have not seen anybody’, but literally ‘I have not seen somebody’)—van der Auwera, Decuypere, and Neuckermans. ).

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

   



() Ewe (ewee, Nada Gbegble p.c.) Ame aɖeke me-fe le abɔ la me o person no -play at garden the in  ‘Nobody played in the garden’ Another difference between Jespersenian doubling and negative concord is that the latter comes in subtypes. Since Giannakidou (: ), one distinguishes between strict and non-strict negative concord. ‘Strict’ means that negative concord is obligatory, and ‘nonstrict’ that negative concord is found when the negative indefinite follows the finite verb, and is impossible when the order is reversed.²² The former is illustrated with Russian, the latter with Chamorro. ()

Russian a. Nikto zdes’ menja ne znajet. nobody here me  knows ‘Nobody knows me here.’ b. Ja zdes’ ne znaju nikogo. I here  know nobody ‘I know nobody here.’

()

Chamorro (cham, Chung : , ) a. ni unu istaba guini gi paingi  one .be here  last.night ‘No one was here last night’ b. ni ma-akka’ yu’ ni háfafa ha’  .-bite I  anything  ‘I wasn’t bitten by anything’

To the extent that one can see from a data set of thirty-four languages in van der Auwera and Van Alsenoy (: ), strict negative concord is more frequent than non-strict negative concord. Perhaps the reason is that strict negative concord is a simple system: the doubling is independent of word order. The non-strict system lacks this simplicity, although it has been argued to be functionally motived too. It is the independently needed Neg-First principle that comes into play (Haspelmath : ). When Chamorro ni unue appears in front of the verb, Neg-First is satisfied and there is no need for a preverbal ni, but when something like ni unu (like ni háfala) follows the verb, the negator comes relatively late and the preverbal no satisfies it. Non-strict negative concord is also more complex in the sense that there are various subtypes. Catalan illustrates one of these: with a preverbal negative indefinite the clausal negator no is optional or better; since the versions

²² This glosses us over that (non-)strictness could be a cline, with negative concord being more or less strict. In Jamaican Creole, for instance, negative concord is overwhelmingly strict for e.g. nobadi ‘nobody’, yet the system ‘leaks’ (van der Auwera and De Lisser ).

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



      

with and without no differ with respect to register, the two versions are semantically equivalent. ()

Catalan (de Swart : ) a. Ningú (no) ha vist Joan nobody  has seen John ‘Nobody has seen John’ b. En Pere no ha fet res the Peter  has done nothing ‘Peter has done nothing’

Or take Georgian. What is relevant in this language, according to King (), is not whether the negative indefinite merely precedes the finite verb, but whether the negative indefinite immediately precedes the finite verb. If that is the case, then the verbal negator is optional. In all other cases, the verbal negator is obligatory. This is illustrated in (). ()

Georgian (King : ) a. šeni cigni versad (ver) vnaxe your book nowhere  .see. ‘I couldn’t see your book anywhere’ b. versad šeni cigni ver vnaxe nowhere your book  .see. ‘I couldn’t see your book anywhere’ c. nu gagzavnit nursed neg .send. nowhere ‘Don’t send it anywhere’

There are also patterns of so-called ‘negative spread’, patterns with two or more negative indefinites with or without a standard negator yielding one semantic negation, as in Korean. () Korean (Hwang : ) Amwuto amwuto-lul mol-n-ta nobody nobody- .know-- ‘Nobody knows anyone’ Based on a data set of some twenty-five languages Zeijlstra (: ) claims that all negative concord languages have negative spread. A final point takes us back to negative quantification, as in (a), repeated as (a) below. The strict vs. non-strict parameter standardly applied to negative concord applies to negative quantification too. To see this, we don’t have to go further than Dutch and English. Dutch illustrates the strict type: the indefinite is negative, there is no standard negator, and the negative indefinite can occur both before and after the finite verb. In English these options also exist but there is a word order dependent alternative and so the

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

   



pattern can be called ‘non-strict’: if the indefinite occurs after the finite verb, it can be a negative polarity item and occur with a standard negator. () Dutch a. Ik geloof niemand I believe nobody ‘I believe nobody’ b. Niemand gelooft mij nobody believes me ‘Nobody believes me’ ()

English a. Nobody believes him b. *Anybody didn’t see me c. I believe nobody d. I don’t see anybody

There are again subtypes of non-strictness (see van der Auwera and Van Alsenoy ). As to frequency, it seems that the non-strict type is more frequent (van der Auwera and Van Alsenoy : ): the strict type only uses the misleading pattern, the one without the standard negator (see top of section ..), whereas the non-strict type allows the standard negator in one of the two word order constellations.

.. P

.................................................................................................................................. This chapter has surveyed recent and ongoing work on the typology of negation, with a focus on standard, prohibitive, and existential negation and on the negation of indefinites. For some phenomena we ventured claims on what is frequent or rare and on why this should be the case. Typology of the last decades of the previous century aimed to relate phenomena with implicational universals (‘if a language has this, then it will also have something else’). This is still a goal of current typology but the implicational approach was not given pride of place here. There are several reasons for this. First, we often simply don’t have enough data to confidently propose an implicational universal. Second, sometimes the claim is trivial and so we didn’t bother the reader with saying that when a language is isolating the negator will not be morphological. Third and most importantly, linguistic reality is a ‘battlefield’ of competing motivations and a matter of tendencies rather than of simple implicational universals of the type that were common twenty to forty years ago. This is most clearly visible in discussions of how the position of negators correlates with other word order properties of a language (Dryer b). Another property of modern typology is the interest in the geography of the phenomena. Our materials did allow some areal statements, but we rarely commented on whether areal convergence is due to contact, to the fact that the area only has genetically related

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



      

languages, or to chance. In some instances contact did play a role, as is clear for the Cham Bahnaric negative doubling discussed in section ... But even if an areal convergence is due to chance, it is good to be aware of the areal dimension. Thus if it is true that negative concord is a typically Eurasian phenomenon (section ..), this will make linguists speaking and studying Eurasian languages prudent in generalizing too much too soon about human language as such.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ......................................................................................................................

 -      ......................................................................................................................

 

.. I

.................................................................................................................................. T chapter deals with the morpho-syntactic aspects relevant to describe the behavior of the negative marker cross-linguistically. It focuses in particular on the plain marker of sentential negation.¹ Descriptively, sentential negation is “a means for converting a sentence S1 into another sentence S2 such that S2 is true whenever S1 is false, and vice versa” (Dahl : ). Sentential (i.e. propositional) negation obtains when the negative operator scopes over the predicative nucleus of the clause. Technically, this means that the negative operator scopes at least over the event variable introduced by the predicate (Acquaviva ; Giannakidou ; Penka : –). The expression of sentential negation is a universal property of languages, which is obtained through various means. This chapter focuses on the expression of sentential negation by means of a negative marker (NM). As customary in cross-linguistic studies on sentential negation, I restrict my attention to the morpho-syntactic aspects of the NM used in declarative main clauses (“standard negation” in Payne ; cf. Miestamo : –). Other dimensions of variation in the typology of negation are treated in Chapter  in this volume.² An NM builds S2 as a syntactically simple sentence, that is it creates no embedding, and its only function is to signal the presence of a negative operator. This definition of the

¹ This chapter will only include markers that can express negation in the propositional domain. For markers of negation in smaller domains cf. Chapters  and . ² The glosses follow the Leipzig Glossing Rules. When examples are cited from external sources, glosses reflect the original as closely as possible. Abbreviations used there and not present in the Leipzig list are:  (connegative);  (final suffix);  (necessity); . (negated existential);  (polarity particle).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

NM includes expressions of the sentential negation operator such as (a) and excludes other expressions of sentential negation like (b), which involves subordination, (c), which additionally expresses quantification over individuals, and (d), which additionally expresses quantification over times. In the course of the discussion we will see that these descriptive criteria, while useful to circumscribe the phenomenon, are sometimes defied by borderline cases (often explained by the fact that b–d can be diachronic sources for NMs).³ ()

a. b. c. d.

John is not running It is false that John is running Nobody is running John is never running

It has been repeatedly remarked in the literature that the forms taken by the NM across languages show wide but also principled variation, and that it is possible to distinguish a number of recurrent forms (Dahl , ; Payne ; Horn ; Kahrel and van den Berg ; Bernini and Ramat ; Zanuttini , ; Miestamo ; Dryer a). The WALS collects , data points on the form of the NM (Dryer a). The results are reproduced in Table ., maintaining the original labels. Under the label “negative particle” both X⁰ and XP markers are subsumed, which I respectively call “negative particles” and “negative adverbs,” as well as negative complementizers. The label “double negation” refers to what I call “discontinuous negation.” The NM can surface in different positions in the clause: these aspects are treated in detail in Chapter  in this volume. In the current chapter we will rather focus on the crosslinguistic diversity in the form of the NM seen in Table ., and discuss positional factors only when clear correlations emerge between them and the form of the NM. It is also

Table .. Cross-linguistic variation in the form of the NM Value Negative affix Negative particle Negative auxiliary verb Negative word, unclear if verb or particle Variation between negative word and affix Double negation 

Representation       ,

Source: Dryer (a).

³ See Chapter  in this volume. Interesting evidence for a diachronic process involving (d) comes from the use of never as a general past negator in the varieties of English investigated by Lucas and Willis ().

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

     



essential to stress that cross-linguistic differences in the (e.g. pre- vs. postverbal) positioning of the sentential NM do not impact on the logical scope of the negative operator, which, in the case of sentential negation, is uniformly above the event expressed by the matrix predicate. This is important evidence for the fact that the NM is a mark for the presence of a negative operator, but it is not necessarily its direct realization, hence its syntactic positioning does not necessarily reflect its scope (cf. Zeijlstra a for extensive discussion). We will see further on (section .) that one way to capture this fact is to assume different featural specifications for the NM, based on the distinction between interpretable [iNeg] and uninterpretable [uNeg] formal features. Across languages the NM typically interacts, both in terms of its form and of its position, with verbal projections (vP, IP / TP): the placement of the NM is determined with respect to the finite verbal element. Therefore, in this chapter I discuss the morpho-syntactic status of the NM in terms of a cline of morpho-syntactic autonomy with respect to the finite verb: I discuss negative affixes (section .), negative particles and adverbs (section .) and cases of discontinuous negation (section .). I furthermore exemplify negative auxiliaries (section .) and negative complementizers (section .), which are borderline cases with respect to the definition of NM adopted here. In the concluding section (section .) I discuss proposed correlations between the morpho-syntactic status of the NM and other syntactic facts.

.. NM  

.................................................................................................................................. The marker of sentential negation can be an inflectional category of the verb. In Swahili it is a prefix preceding all the other inflectional morphemes (a); in Turkish it is a suffix following the verbal root (and valency-changing suffixes if present) but preceding tense endings (b); in Paez the negative suffix follows aspectual suffixes but precedes the subject marker (c). ()

a. Swahili (Devos and van der Auwera : ) ha-tu-ta-lim-a ---cultivate- ‘We will not cultivate’ b. Turkish (Kornfilt : ) Hasan kitab-ı oku-ma-dı Hasan book- read-- ‘Hasan didn’t read the book’ c. Paez (Dahl : ) u’x-we-ts-meː-th go----. ‘I don’t go / I’m not going’

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

Dahl () calls this strategy “morphological negation,” and reports that it is the most frequently attested one in his sample of approximately  languages (%). Miestamo (: ) groups this strategy together with other non-inflecting NMs, such as clitic and non-bound particles (section .), and finds bound negative morphemes in % of the languages in his sample. Dahl () also remarks that languages in his database that display the morphological strategy choose suffixation more frequently than prefixation, in accordance with Greenberg’s (: ) observation that inflectional phenomena prefer suffixation to prefixation.⁴,⁵ This observation is confirmed by Miestamo’s larger, more balanced sample. Dahl (: –) comments on the variation observed in the order with respect to other inflectional morphemes (cf. ()). He also observes that, in about half of the languages with affixal NM in Miestamo (), the NM is the outermost morpheme in the word; in particular, negative prefixes are almost always word-initial. Suffixal negation is more integrated in the verbal morphology, can undergo fusion with other morphological categories, and can appear between other inflectional suffixes (cf. (c)). The typological evaluation of these phenomena is complicated by the difficulty of empirically distinguishing between affixes and independent morphemes, both in outermost pre- and postverbal position. Especially in the case of prefixes, it is sometimes difficult to decide whether the NM is an affix or a clitic particle (section .). The clearest cases of affixal NMs are those inserted amid other affixes, and those where morphological fusion with other verbal categories (tense, subject agreement) is observed. The distribution of affixal NMs is consistent with hypotheses that analyze them as heads of a dedicated projection NegP within the inflectional domain (Pollock’s  “Split INFL” hypothesis). Various implementations are possible depending on the theory of the morphology-syntax interface adopted (cf. Zeijlstra :  for discussion). An aspect requiring further research concerns the interaction of affixal negation marking with tone reported for some Niger-Congo languages (Dahl : –, : , and references therein). Igbo is a particularly interesting case: according to Obiamalu’s () analysis, negation is jointly marked by the suffix -ghi and the high tone (a prosodic morpheme) on a c-commanding element (e.g. the AGR prefix e- in (), or an auxiliary verb). ()

Igbo (Obiamalu : ) Ezè e-rī-ghī nri Eze -eat- food ‘Eze did not eat food’

Optionally, the -ghi suffix may be missing and negation may be expressed by high tone alone (Obiamalu : –).

⁴ Cf. Chapter  in this volume for negative affixes involved in derivational operations instead, such as English un-, Italian in-. ⁵ A rare case of reduplication of part of the verb stem to express negation, Tabasaran, is mentioned by Dahl (: ).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

     



.. NM      

.................................................................................................................................. Marking sentential negation by means of free non-inflecting morphemes is another very frequent strategy cross-linguistically. ()

a. Italian Gianni non ha risposto Gianni  have- answer-. b. German Hans hat nicht geantwortet Hans have-  answer-. ‘John hasn’t answered’

On the basis of distributional factors and of the interaction with other aspects of negation systems, formal syntactic research has clearly shown the need to distinguish between negative particles (a) and negative adverbs (b). The distinction is motivated in phrasestructural terms: negative particles are syntactic heads (X⁰) projecting their own functional projection NegP; negative adverbs are maximal projections (XPs), occupying the specifier position of NegP or adjoining to verbal projections. Research stemming from Pollock’s () foundational work has established that negative particles and adverbs behave differently from other apparently similar elements (pronominal clitics, non-negative adverbs): this has led to the postulation of a dedicated functional category, NegP. The ensuing debate has revolved around the position of this category in the clausal spine: influential ideas include Ouhalla’s () NEG Parameter, Laka’s () proposal of a high projection for polarity, Zanuttini’s (, ) model with various NegP projections in the clause, Zeijlstra’s (a, ) Bare Phrase Structure approach, Poletto’s (, ) Big NegP hypothesis (cf. below and Chapter  for discussion). Zanuttini’s (, ) work on Italo-Romance provided the first systematic observations on cross-linguistic variation depending on the NM’s phrase-structural nature. Haegeman () and Zeijlstra (a) integrated Germanic data into the picture. The classical tests for establishing the X⁰ / XP nature of the NM are based on the interaction with verb movement and on the behavior in isolation or in adjunction structures (cf. Zanuttini , ; Zeijlstra a: – for an overview). X⁰-NMs interact with verb movement, blocking for example V-to-C movement in questions and imperatives as a consequence of the Head Movement Constraint. For instance, Zanuttini (: –; : –) shows that in Paduan the NM blocks V-toC in questions (where V-to-C is evidenced by subject-verb inversion) (), whereas in Swedish the NM does not block V-to-C (verb second) in declarative matrix clauses (): ()

Paduan (Zanuttini : , ) a. Vien-lo? come-.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



  b. * No vien-lo?  come-.

()

Swedish (Zanuttini : ) inte köpte boken a. . . . om Jan . . . that John  bought books ‘ . . . if John didn’t buy books’ b. Jan köpte inte boken John bought  books ‘John didn’t buy books’

In languages with X⁰-NMs, verb movement can happen only if the NM can cliticize to the verb and moves along with it. XP-NMs do not block verb movement, but interact with A′ phrasal movement. According to these tests, clear cases of X⁰-NMs are Romance and Slavic preverbal NMs (e.g. Italian non, Spanish no, French ne, Russian ne, Polish nie); clear cases of XP-NMs are lower NMs (occurring between the finite auxiliary and the participle in structures like b) like French pas, Swedish inte, Dutch niet, German nicht. Merchant’s () ‘why not?’-test singles out XP-NMs since X⁰-NMs would not be able to adjoin to a maximal projection. ()

a. Italian * perché why

non? not

b. German warum nicht? why not ‘Why not?’ Merchant () assumes that ‘why not’-structures originate through adjunction of the NM to the wh-element; however, the test remains valid also assuming a derivation through ellipsis following XP-movement of the NM from a lower merge position to a higher polarity or focus projection (cf. the derivation of short answers in Holmberg ). The test cannot be applied if the language has a homophonous negative answer particle, which is a maximal projection, since it would be impossible to ascertain whether the adjoining item is the NM or the answer particle. More in general, the cross-linguistic applicability of the tests seen above is problematic (and raises related acquisitional issues), since “most of them require a particular constellation of properties in order to be applicable in a given language” (Zanuttini : ). For instance, movement tests easily individuate as XPs NMs that may surface postverbally. However, things are more difficult if, for some independent reasons, the verb never raises past the NM in a language. In this case, other clues, such as the ‘why not’-test, the possibility to displace the NM for information-structural reasons, and language-internal considerations relating to word order may be more revealing.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

     



An instructive example of the intricacies involved in establishing the X⁰ / XP nature of the NM comes from Hindi. The Hindi NM nahiiN (compound of negative particle na and emphatic particle hiiN ‘only’) surfaces between the object and the verb when expressing sentential negation, as in () (Mahajan ; Dwivedi ; Kumar ): ()

Hindi (Kumar : ) maiN aam nahiiN khaa-taa I mango  eat- ‘I do not eat mango’

When used as constituent negation instead, it always follows the focused constituent: ()

Hindi (Kumar : ) maiN aam khaa-taa nahiiN I mango eat-  ‘I do not eat mango’ [context: I do not eat mango, but I sell mango]

As constituent negation, nahiiN is analyzed as a X⁰ taking the focused constituent as its complement and surfacing to its right, in conformity to the head-final nature of Hindi (Dwivedi ). As sentential negation, however, a X⁰-analysis (head of a NegP between AspP and TP, as in Kumar ) is problematic in view of word order, since the NM precedes the inflected verb. To solve this issue Sheehan () proposes that nahiiN has a double status: as constituent negation it is a X⁰ but as sentential negation it is a XP in the specifier of a NegP immediately dominating VP. The idea that NMs can have variable phrase-structural status in the same language is also defended in Matushansky (: –), who draws a comparison with pronominal clitics. It is further discussed by Breitbarth () in the context of historical processes targeting the NM (especially Jespersen’s Cycle, cf. Chapter  in this volume) and leading to structure minimization. Breitbarth (: –) applies to NMs Cardinaletti and Starke’s () notion of structural deficiency of pronominal forms (cf. also Cardinaletti  for an extension to adverbs). She proposes that NMs have a complex internal structure, which is subject to variation and change: strong adverbs have a C layer, corresponding to focus; weak adverbs lack C but have a Σ layer, encoding prosodic information; clitic forms lack this layer and only project the I layer, containing negative features; affixal adverbs are simple Adv projections. ()

variable structure of NM (adapted from Breitbarth : ) [CAdvP [ΣAdvP [IAdvP [AdvP ]]]]

Depending on the amount of internal structure, NMs have different distributional properties. In some languages the same form may realize different structures: English not may realize ΣAdvP when used as plain sentential negation or CAdvP when used as stressed contrastive particle; the form n’t would instead correspond to the IAdvP structure (on the

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

dual status of the English NM cf. already Haegeman ; Zeijlstra a: –; Matushansky : –).⁶ Similarly, one could argue that Spanish no projects less structure when it is a NM that can cliticize to the verb and move along with it (similar to the analysis of English n’t given by Matushansky ), but has a stronger variant with more internal structure when it is used as negative answer particle (and in ‘why not’-structures like ¿porque no?). As Breitbarth () remarks, her proposal is similar to Poletto’s (, ) big NegP hypothesis (on which cf. Chapter  in this volume) in positing a complex internal structure for NMs. However, while in Breitbarth’s proposal NMs attach to the targeted constituent via external Merge, Poletto assumes one invariable position for the first merge of the NMs and subsequent movements of subparts of the structure to higher positions (for a movement-based approach to the syntax of Italian non cf. Belletti , ). A further issue concerns the distinction between subtypes of X⁰-NMs. Preverbal NMs of Slavic and Romance have been analyzed as negative heads, hosted in the dedicated projection NegP closing off the inflectional domain. However, Romance NMs have a more autonomous status, and can be separated from the verb by pronominal clitics; Slavic NMs instead obligatorily cliticize to the finite verb, and nothing may intervene. This in turn connects to the difficulty, mentioned in section ., of distinguishing non-bound X⁰-NMs from bound prefixal NMs. A much-discussed case is the contrast between Polish, where the NM is a free morpheme, and Czech, where the cognate NM is a bound affix: () Polish (Miestamo : ) nie czyta-m  read- ‘I don’t read’ ()

Czech (Miestamo : ) ne-vol-al -call-. ‘He did not call’

The behavior of the NM in Czech is reminiscent of Persian, where the NM (which continues the same proto-Indoeuropean NM *ne) is analyzed as a prefix: () Persian (Kahnemuyipour : ) Ali in ketaab-o na-xarid Ali this book- -bought ‘Ali didn’t buy this book’ The main criterion followed by grammatical descriptions is orthographic (Dahl : , : ; Miestamo : –); as Dahl (:  fn. ) remarks, in the case of (–) ⁶ See Biberauer () for another set of arguments leading to similar conclusions on the existence of homophonous, yet semantically and structurally distinct, negative particles in languages such as Afrikaans and Brazilian Portuguese.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

     



orthography seems to reflect the fact that in Czech there is indeed a stronger prosodic unit between the negative particle and the verb, due to the language’s initial stress rule. However, orthography is obviously not always reliable: in Serbian/Croatian, the negative particle ne cliticizes to the first finite form (auxiliary or lexical verb), (a); with a few verbs (‘have’, ‘be’, ‘want’), this is mirrored by orthographic conventions (b). Nonetheless, the particle’s syntactic properties are invariant: ne is always strictly adjacent to the verb and cannot be separated from it by pronominal clitics (unlike e.g. Italian non).⁷ () Serbian/Croatian (Progovac : –) a. Milan ne poznaje Mariju Milan  knows Mary ‘Milan does not know Mary’ b. Milan neće pobeći Milan -will run-away ‘Milan will not run away’ The difference between clitic and non-clitic X⁰ has often been interpreted in diachronic terms as instantiating two sub-stages of Jespersen’s Cycle, with the clitic form representing the stage potentially triggering reinforcement (van Gelderen ; Jäger ; Breitbarth ). In turn, adverbial NMs often have a more recent historical origin and may occupy a low syntactic position, below the inflected verb, consistent with their frequent origin as argumental negation strengtheners. In other cases, negation strengtheners bleaching to plain NMs may be found high in the clause, as manco in the Southern Italian variety studied by Garzonio and Poletto () and mica in Northern Italian dialects (Cinque ; Penello and Pescarini ). Zanuttini () and Poletto () have shown how the NM’s etymological origin may influence its synchronic distribution (cf. Chapter  in this volume).

.. D 

.................................................................................................................................. Discontinuous negation is observed when two morphologically and positionally distinct elements obligatorily co-occur in the marking of sentential negation. Discontinuous negation is a combination of the strategies seen in section . and section ., and is best analyzed as an intermediate stage of Jespersen’s Cycle (Chapter  in this volume). The original negation, as well as the reinforcing element, may be a bound affix or a free particle. Reinforcers may express a multiplicity of emphatic meanings, resulting for example in scalar focusing of negation, intensification, or rejection of a presupposition (cf. Chapter  in this volume).

⁷ Serbian/Croatian ne can also be used as constituent negation, and in that case it carries its own accent (Alexander : , –). It seems that also in this case an approach that allows for structural variability within one language fares better than an explanation in terms of homonymy.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

() Guaraní (Miestamo : ) n o-hò i šẽ´ne  -go  .. ‘He will not go’ Bantu languages often show the doubling of a verb-related negative affix by means of various kinds of historically more recent reinforcers, which can take the form of affixes or of independent particles (Devos and van der Auwera ): () Beembe (Devos and van der Auwera : ) me n-síí-tin-a mukáándá ko I --write- .letter  ‘I have not written a letter’ Pollock (: ) proposed for French ne . . . pas that both the head and the specifier of NegP can be filled, and that the head can move to cliticize to the verb in TP, yielding discontinuous negation in French: ()

[TP nei [NegP pas [Neg⁰ nei] [VP . . . ]]]

This analysis has been extended to other languages exhibiting discontinuous negation. In van Gelderen (: –) doo and da of Navajo () are analyzed as, respectively, the specifier and the head of a NegP dominating the VP; doo is a newer element diachronically emerging as reinforcer. () Navajo (Young and Morgan : ) doo dichin nishłí˛į da  hungry .be  ‘I am not hungry’ In general, the discontinuous negation patterns widely attested in Athabaskan languages are considered by van Gelderen to be the result of functionally similar diachronic cycles (although the reinforcer has different origins in the various languages). Similar accounts have been proposed for discontinuous negation in Afro-Asiatic languages (cf. van Gelderen : – for an overview; Ouhalla  and Ouali  for Berber; Fassi Fehri : –; Shlonsky : – for Hebrew; Shlonsky : – and Lucas  for Arabic; for historical aspects of Semitic negation cf. also Sjörs ). ()

Berber (Ouali : ) ur ughax sha lktaab  -bought-  book ‘I did not buy the book’

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

     



Dahl (: ) remarks that “it seems to hold that the two particles appear each on one side of the F[inite]E[lement].” This tendency could be related to the fact that, in Jespersen’s Cycle, preverbal reinforcement of the NM often results in the substitution of the NM itself through an emphatic variant, without a doubling stage, cf. Rionerese manco (Garzonio and Poletto ) and Latin noenum (Gianollo : –). When the doubling is optional, as in Welsh, it conveys an emphatic meaning: () Welsh (Borsley and Jones : ) Dw i ddim wedi cysgu dim be.. I   sleep  ‘I haven’t slept at all’ Research on languages with optional doubling has shown that it is doubtful that reinforcing elements are always the realization of the specifier of NegP. Biberauer () shows that not all sentence-final negative particles, comprising the doubling ones, can receive the same analysis in each language. For the Afrikaans concord element nie, surfacing in phrase- or sentence-final position, she proposes that it is a PF-realization of a Pol(arity) head valued through Agree when a negative element (the linearly first nie) is inserted in the structure. Also the Brazilian Portuguese emphatic construction with doubled não () is accounted for as spell-out of an agreement relation. ()

Brazilian Portuguese (Biberauer and Cyrino : ) comprou a casa (não2) Ele não1 he  bought the house  ‘He has not/NOT bought the house’

Martins () provides a comparative analysis of this and similar Romance constructions, arguing in favor of locating the doubling element in a left-peripheral position (CP), and explaining the sentence-final position by means of syntactic inversion.

.. N 

.................................................................................................................................. NMs with the form of negative auxiliaries (“inflecting” NMs in Miestamo’s :  terminology) are known from various languages: they are particularly frequent in Northern Eurasia (Dahl : ), for example in the Uralic languages discussed in Miestamo (), van Gelderen (, ), Hamari (); other examples include Classical Japanese (Kato ), the ‘long’ construction of Korean (Miestamo : –), as well as ain’t in varieties of English (Cheshire ).⁸

⁸ Less frequently, a non-negative auxiliary is used in negative constructions, as opposed to auxiliaryless positive ones (Miestamo : –; Dahl : –).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

Finnish (–) marks sentential negation by means of a negative auxiliary inflected for subject agreement, whereas T(ense)A(spect)M(ood)-categories are realized on the nonfinite form of the verb: ()

Finnish (Dahl : ) a. lue-n read- ‘I read’ b. e-n lue - read. ‘I do not read’ c. lue-t read- ‘You read’ d. e-t lue - read. ‘You do not read’

() Finnish (Miestamo : ) e-n syö-nyt omena-a - eat-. apple- ‘I didn’t eat an / the apple’ The verb is in a non-finite form (the connegative in the present and the perfect, the past participle in the past and the pluperfect, Miestamo : ). In Evenki the negative auxiliary e- is inflected for TAM and agreement and is followed by a participial form of the lexical verb: () Evenki (Nedyalkov : ; Miestamo : ) nuŋan min-du purta-va e-che-n bū-re he - knife- -- give- ‘He did not give me the knife’ Syntactic analyses of negative auxiliaries often involve movement operations from lower Neg positions to higher clausal projections, where they value their inflectional features (cf. van Gelderen ). Uralic negative auxiliaries historically develop into negative particles in Mansi and Khanty (Kiss : ), pointing to a recurring diachronic path (Honti ; van Gelderen ).⁹ According to van Gelderen () this form of the negative cycle is motivated by structure minimization, similarly to Jespersen’s Cycle.

⁹ In Hungarian the negative auxiliary has been substituted by the negative particle nem, originating from an indefinite pronoun (Kiss : ).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

     



In some cases it is difficult to distinguish a negative auxiliary from a negative adverb. This happens when the verb in the language does not carry inflectional morphology; as a consequence, neither does the negative auxiliary, becoming morphologically indistinguishable from an adverbial NM. A case in point is Maori, where one must resort to syntactic evidence in order to prove the verbal nature of the NM kaahore (Pearce ): () Maori (Dryer a) kaahore taatou e haere ana aapoopoo  . / move / tomorrow ‘We are not going tomorrow’ Biberauer (: –) discusses the case of Ma’di, where the NM takes two different forms, depending on tense: () Ma’di (Blackings and Fabb : –) dʒótī kʊ̄rʊ̀ a. ḿ-āwí -open door . ‘I did not open the door’ kʊ̄ b. ıd́ rɛ́ ɔ̄-ɲā ı ̀zá rat -eat meat . ‘Rats don’t eat meat’ Despite the appearance, however, these forms are not inflected auxiliaries (i.e. they do not instantiate the realization of a T node), but are better analyzed as two different negative particles sensitive to tense, which in Ma’di is encoded compositionally by means of various devices (tone, inflection or absence thereof, final particles). In some languages, special negative auxiliaries exist only for expressing negative existential predication, that is existential forms, but also copulas, are negated with different formal means than other verbs in the language.¹⁰ In Turkish the negative copula degˇil is used when the predicate is of nominal nature. The copula can carry inflection for agreement, tense, mood: () Turkish (Kornfilt : ) (ben) hasta degˇ il-im I sick .- ‘I am not sick’

¹⁰ Also Welsh has special third person forms of ‘to be’ under negation, e.g. maent ‘are’, ydynt ‘aren’t’. However, Welsh is special in featuring different negative forms for many verbs, due to the morphophonological effects (mutations) of the preceding original NM ni(d), which is retained in formal Welsh but has disappeared from the informal variety (Borsley and Jones : ch. ; Willis a). Paradigmatic effects explain the existence of special negative forms of the verb be in varieties of English, e.g. the use of weren’t for all persons (Britain ).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

Hungarian shows a special form for the negated existential predicate, nincs, in the present tense (a), alongside the regular negative construction in the past tense (b). ()

Hungarian (Croft : ) a. itt nincs taxi here .. taxi ‘There’s no taxi here’ b. nem volt sok ház a kis uccá-ban not .. many house  little street- ‘There weren’t many houses on the little street’

A recent case of grammaticalization of a new negative existential is discussed by Willis (a): southern Welsh smo derives from the univerbation of NM nid, rd sing. form of ‘be’ oes, and mark for negated definite object / unaccusative subject mo: () Present-day southern Welsh (Willis a: –) Smo ’r gath ’ma. .. the cat here ‘The cat isn’t here’ Croft () shows how these special negative existential forms may diachronically evolve into the marker of standard sentential negation (“Croft’s Cycle,” cf. also Veselinova ). For instance, one of the NMs in Chinese, mei, derives from a negative existential verb, meaning ‘to die’ in Old Chinese (van Gelderen : –). A similar development involves negative clefts and often results in pragmatically marked NMs. Jewish Babylonian Aramaic lāw, used in denials and in counterfactuals, derives from the univerbation of the standard NM lā and the agreement marker -hu of the elided copula with which it combined in clefts (‘it is not the case that . . . ’) (Bar-Asher Siegal and De Clercq ). Similarly, the presuppositional negation neca found in the Sicilian variety of Mussomeli (Cruschina ) originates from a cleft construction (un jè ca, ‘not it.is that’). Differently from negative auxiliaries, these latter cases, however, involve a former clause boundary, which disappears in the grammaticalization process. In fact, a few languages attest marking of plain sentential negation by means of a superordinate verb (Payne : –). Since superordinate negative verbs create subordination, they do not strictly adhere to the definition of sentential negation adopted here. As Miestamo (: ) notes, it is not always easy to empirically distinguish between negative auxiliaries and higher verbs. In Tongan the negated existential form ’ikai (.a) is also used as the standard negation in declarative clauses; according to Payne (: ), Miestamo (: ), in (b) ’ikai is a superordinate verb taking a complement clause. () Tongan (Churchward : ; Croft : ) a. ’Oku ’ikai ha faiako ’i heni  . a teacher at here ‘There isn’t a teacher here’

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

     



b. Na’e ’ikai [ke ’alu ’a Siale ]  .()  go  Charlie ‘Charlie didn’t go’ Also in Squamish the NM haˈṷ is analyzed as a higher predicate complemented by a subordinate clause (introduced by the irrealis marker q-): () Squamish (Miestamo : ) haˈṷ q-ʔan-cỉcẚṕ’  --work ‘I don’t / didn’t work’ The use of superordinate verbs as NMs is also observed in languages that have a separate form for negation in prohibitives and other non-declarative constructions (Latin prohibitive noli / nolite, the imperative forms, respectively second person singular and plural, of nolo ‘not want’; Welsh peidio, a special negative non-finite verb negating non-finite clauses and imperatives), cf. further Croft ().

.. N 

.................................................................................................................................. The expression of sentential negation by means of complementizers is rare in declarative main clauses, to which this chapter is limited. Negative complementizers have special properties in terms of sensitivity to modality, and, possibly due to their clause-typing abilities, they are more frequently expressions of non-standard negation (e.g. in prohibitions, cf. Horn : –). Hausa has a negative complementizer for sentential negation in main declarative clauses, which also expresses tense, aspect, and person features; in Dryer (a) it is not treated as a negative auxiliary because the verbs in the language do not inflect for the categories expressed on the complementizer. () Hausa (Dryer a) bàn ga yārò-n dà bài tàimàki Lādì ba .. see boy-  ... help Ladi  ‘I didn’t see the boy who didn’t help Ladi’ Irish is a well-known case where standard negation in declarative main clauses is expressed on C, linearly preceding both the verb and its subject (McCloskey , ). () Irish (McCloskey : ) Ní chuireann sé isteach ar phostanna - put. he in on jobs ‘He doesn’t apply for jobs’

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

Irish has a full system of complementizers with interrogative and negative forms. Negative complementizers in finite clauses include ní in main clauses, and nach in embedded ones (different forms are found with past tense). McCloskey analyzes structures like () as instantiating an Agree relation between a lower Pol(arity) projection and the higher C node, whereas Duffield () argues for a movement analysis. Also for Moscati (), negation in C is the result of a raising operation (from NegP to C). Roberts (: –) and Biberauer and Roberts () propose an alternative licensing mechanism based on Ouali’s () Keep-Donate-Share system: they treat negation as a clause-typing feature in C, which may be kept, donated, or shared with the lower T category, yielding different syntactic realizations. Irish would be a case where negation is kept in C. Treating negation as a clause-typing feature seems a promising way to unify these cases with those of negative complementizers introducing prohibitives and subordinate clauses (e.g. Ancient Greek mē, Chatzopoulou ; Modern Greek mi(n), Roussou ; Latin ne, Orlandini ; Southern Calabrian nommu, Damonte ; Arbëresh mɔs, Manzini and Savoia ; Welsh na, Roberts : –).

.. C      

.................................................................................................................................. In this section I briefly discuss two areas in which the shape of the NM has been argued to interact with other syntactic aspects: tendencies in word order (on which cf. Chapter  in this volume) and the relation between the form of the NM and Negative Concord (on which cf. Chapters  and ). Word order tendencies for NMs have been discussed in terms of either a ‘special’ or a ‘harmonic’ position of the NM with respect to other clausal elements. The possibility that NMs may follow special positioning rules has been discussed under the heading of the Neg-First Principle, defined by Horn (: ) as “the preference for negation to precede its focus.” Horn attributes the principle to Jespersen (: ), who says that negative elements are placed “first, or at any rate as soon as possible” in the clause. The principle has been integrated into de Swart’s () OT treatment of the syntax of negation based on the ranking of the grammatical constraints NegFirst and FocusLast. As Dahl (, ) remarks, the principle may in fact be decomposed into a number of separate claims. Dahl’s () survey does not support the strictest interpretation of the Neg-First Principle, that is that the NM tends to appear sentence-initially or as close as possible to the sentence-initial position. Rather, the position of the NM is tied to the position of the finite verb (Dahl : –; : –), and a more sentence-initial positioning of the NM is a consequence of a more sentence-initial positioning of the finite verb itself. According to Dahl (: ), –% of the cases discussed by Dryer () show the NM adjacent to the verb, either before or after it. Negative auxiliaries behave like other auxiliaries in the language, that is their positioning is determined by the other inflectional categories expressed on them. In general, the positioning of NMs appears to

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

     



be fixed, also in languages with more flexible word order possibilities (although in some languages it may be more sensitive to focus, cf. Chapter  in this volume). Hypotheses that the order of the NM may be harmonic with other categories in a language (Bartsch and Vennemann ; Lehmann ) have been falsified by crosslinguistic research. In Dahl’s () database the preverbal position appears to be preferred for non-inflectional NMs, independently of the position of the object. In general, no correlation with head-directionality in other categories emerges (cf. the discussion in Dahl : –; : –). Indeed, a harmonic positioning of negation is implausible in view of the acategorial status of the NM highlighted by Cinque (: –) and Biberauer, Holmberg, and Roberts (: –, ). A conclusion that, instead, is corroborated by cross-linguistic evidence concerns the correlation between the phrase-structural status of the NM and the presence of Negative Concord in the language. Also this observation goes back to Jespersen (), who notes the connection between the reduced morpho-phonological substance of the NM and the tendency for it to be doubled by other elements (adverbs or arguments). Formal research (Zanuttini ; Haegeman ; Rowlett ) has equated the weakness of the NM with its phrase-structural status as a X⁰, and has causally connected the presence of an X⁰-NM to the presence of Negative Concord (). () Serbian/Croatian (Progovac : ) Milan *(ne) vidi ništa Milan  sees nothing ‘Milan cannot see anything’ In Zeijlstra’s (a) implementation, the fact that languages with X⁰-NMs exhibit Negative Concord has been attributed to the presence in the derivation of a syntactic projection NegP, which would instead be absent in so-called Double Negation languages, where no Negative Concord obtains. NegP is the projection of the X⁰-NM. In Zeijlstra (a) the postulation of NegP is motivated (also in acquisitional terms) by the existence of syntactic dependencies between it and other items in the clause, such as the indefinite ništa in (). The syntactic dependency, Negative Concord, is interpreted as a form of agreement, triggered by the presence of uninterpretable formal negative features [uNeg] on the indefinite (Reverse Agree). The correlation between the phrase-structural status of the NM and the presence of Negative Concord is unidirectional: a X⁰-NM is always accompanied by (strict or non-strict) Negative Concord; a XP-NM, instead, can be found either in a Negative Concord or in a Double Negation language. This is due to different Merge possibilities for a NM that is a maximal projection: according to Zeijlstra (a), the NM occupies the specifier of NegP only if in the language there is evidence for a syntactic dependency between featurally negative items (e.g. Bavarian ned). Otherwise, no NegP is postulated during acquisition, and the XP-NM adjoins to a verbal projection (e.g. German nicht, cf. also Matushansky : – for English not). In Zeijlstra’s (a, ) system, X⁰-NMs are further distinguished on the basis of their featural content: they can bear an interpretable formal feature [iNeg], or an uninterpretable formal feature [uNeg]. In the latter case, the NM does not directly introduce the logical operator of negation, but needs to be licensed by a higher covert operator. This derives the

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

strict vs. non-strict distinction in Negative Concord systems (cf. Chapter  in this volume): the NM of non-strict systems like Italian carries an interpretable [iNeg] feature, whereas the NM of strict systems carries an uninterpretable [uNeg] feature. Diachronic hypotheses such as van Gelderen (: ; : ) interpret the difference between [iNeg] and [uNeg] NMs as connected to the form of the NM and propose a possible historical cline, within Jespersen’s Cycle, leading from non-bound [iNeg] X⁰-NMs to bound (affixal or clitic) [uNeg] X⁰-NMs. However, cases like Modern Greek dhen, a non-bound NM in a strict Negative Concord system, appear to weaken the supposed correlation between featural import and phrase-structural status, and call for further research integrating the theoretical and the diachronic perspective. In conclusion, this chapter has shown that the expression of sentential negation by means of NMs is characterized by remarkable cross-linguistic variation, but also by a number of organizational principles that have allowed typological and formal research to detect patterns and regularities, concerning both form and distribution. The morphosyntactic properties of NMs interact in interesting ways with other aspects of the syntax of negation. These interactions represent fundamental evidence for current models of the syntax-semantics interface, but at the same time raise issues for future research. The most evident issues that have emerged in this chapter concern the best way to capture the phrasestructural status of NMs and, consequently, their relationship with other constituents, as well as the role of their featural specification in governing their distribution.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

  ......................................................................................................................

                       ......................................................................................................................

 

.. I

.................................................................................................................................. S negation is exceptional with respect to other functional elements universally found in all languages across the world, since it does not always occupy the same position in the clause. While generally elements (either auxiliaries, adverbs, or even bound morphemes) expressing semantic values related to the utterance like (different types of ) modality, tense, (different types of ) aspect etc. can be shown to occupy the same position in the structural tree in all languages of the world (see Cinque  and subsequent work in cartography on this), sentential negation can be expressed in different languages at all levels of clausal structure, ranging from the CP to an IP-internal position dominating TP, to lower positions in the aspectual field, or even in sentence final position. Before asking the question as to why this is so, we have to determine whether sentential negation can indeed appear interspersed with any functional projection in a cartographic clausal spine, or whether there is only a subset of positions where it can surface. In this work I will discuss the possible positions where a negative marker can be found in different languages mainly basing my evidence on non-standard Romance varieties; in particular I will use data from Italian dialects,¹ which display a wide spectrum of negative markers that is similar to the

¹ Traditional dialectologists (see e.g. Rohlfs ) refer to the variation found in Italian dialects as the “small Romania,” since the span of variation found in Italian dialects is similar to the one found in the whole Romance linguistic domain. The use of Italian dialects also justifies itself in the sense discussed by Kayne (): a cohesive linguistic domain allows us to see linguistic phenomena under a magnifying lens and make our observations as similar as possible to a controlled scientific experiment where other variables except the one we are investigating are controlled for. This is so because the grammar of genetically closely related dialects is more similar than that of languages that are very different both from the typological and from the genetic point of view.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

one found in typological work and for which we can trace back the etymological origin. I will also make occasional reference to other language groups that manifest a similar distribution. I will consider both standard sentential negative markers using the definition of Miestamo (: ), that is “the basic way(s) a language has for negating declarative verbal main clauses” as well as non-standard ones, that is those that in addition to negating the clause also add either a presupposition or a sentence implicature to the negative value. Non-standard negations are also those negative markers that are used to express focus on the negative marker itself; see van der Auwera, Vossen, and Devos () for a discussion in typological terms.² The chapter is organized as follows: in section ., I describe the positions that have already been identified in the literature and provide empirical tests to distinguish them. I also discuss the relation between the position of the sentential negative marker and its etymological origin, a fact that is pretty evident in a microcomparative perspective, although once we have identified the possible basic etymological types, the same tendencies are also evident in macro-comparative work. Section . deals with a rather fundamental question, that is whether the positions where we see the negative marker are first-merge positions or whether all negative markers are first-merged in a rather low position (e.g. one located at the border of the vP) and then some of them are moved to satisfy other features the element representing negation intrinsically possesses. Section . deals with the problem of the position of non-standard negative markers and shows that there is no connection between the status (as clitic or independent XP) or the (CP, TP, Asp, or vP) position and the standard versus non-standard interpretation of the negative marker.

.. P   

.................................................................................................................................. Since the seminal work done on Northern Italian dialects by Zanuttini (), we know that negative markers can occupy different positions. Zanuttini identifies a position higher than TP in the clitic field (and some negative markers interact with clitics, in the sense that the clitic disappears when the negative marker occurs, as is the case in Friulian) or as an independent head and argues against the clitic status of standard Italian non on the basis of two arguments: (a) non never occurs in enclisis as a pronominal clitic; (b) non can bear word stress, while clitics cannot. She proposes that so-called “preverbal” negative markers, or better TP negations, can thus be split into two types, those that have a clitic status (like French ne, as analyzed by Pollock ) and those that do not (like standard Italian non).³ Notice however that under the assumptions that all clitics head their own projection in the clitic field, we can simply consider the two types of preverbal negative markers as heading

² For a discussion on the definition of standard negation see Miestamo () and van der Auwera, Vossen, and Devos (). ³ Garzonio and Poletto () argue for the idea that preverbal negative markers can become clitic and give rise to the Jespersen cycle only when several factors are met. One of them is the bi-morphemic status of the negative marker.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

    



two distinct projections both located within the field of clitic positions, one higher and one lower than object clitics. This in turn means that the distinction between the two types of preverbal negative markers is not a clitic/tonic distinction but a distinction in terms of height of the projection.⁴ Zanuttini () also identifies three further positions: The second position to be considered is the one above aspectual adverbs like ‘already/not.yet’. Although Zanuttini does not make any reference to etymology, this position is the one typically occupied by a minimizer category, of the French/Italian pas/mica type. The third position Zanuttini identifies is also located in the lower IP area (as defined by Cinque ) lower than ‘already/not.yet’ but higher than perfective aspect and adverbs like ‘always’. Again, a link to the etymological origin of the negative marker can be established, since the majority of the dialects that use this position have a negative marker that typically surfaces with the form of the n-word corresponding to ‘nothing’ which is the etymological origin of the German nicht, English not, and Occitan ren, Raetoromance nia. One further position for negative markers is in the VP area as a modifier of the object (like German kein). The last type of negative marker Zanuttini identifies is the sentence final Focus one (like Afrikaans nie, Brasilian Portuguese nao, and Milanese no) and proposes the following sentence structure: () [NegP non [TP V+Agr [NegP mica [ TP [AdvP already] [NegP niente [ Asp perf. Vpast part [Asp gen/progr [AdvP always] [NegP NO]]]]]]]] In Zanuttini’s survey there are two further types of negation missing. One is found in the CP area, often expressed by negative complementizers (such as Latin ne or the fully-fledged system of Irish negative C elements ni, cen, etc.). Some Sicilian varieties also have a C negation (see the forms can and nommu, pemmu discussed in Garzonio and Poletto b). The other fundamental type missing are the negative determiners like German kein-, which is clearly integrated in the argumental structure of the verb, that is inside the vP. The respective position of verb and negation is not a good indicator of different negations, since we know that the verb can move to different positions even in varieties that are very closely related to the ones we are dealing with. Since Cinque () we know that adverbs are a much more reliable test, as they generally do not move (apart from focus, which can easily be detected, or lexical ambiguity, which can easily be eliminated by performing the test in more than one language). Apart from the position with respect to adverbs, there are other tests that can be used in Italian dialects to show the necessity of postulating more than one position for negation. I summarize in Table . the tests and the results provided by the negative markers considered by Zanuttini () and systematize them for all negative marker types. Table . shows that the four NegPs identified by Zanuttini behave differently not only with respect to their position in relation to adverbs (as the first row illustrates), but also

⁴ I refer here to Zanuttini () for the examples of negative markers that interact either with subject or with object clitics.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

Table .. Properties of different types of negative markers Position V to C interference Negative concord Compatible with true imperatives Reorders with clitics

CP Neg

NegP NegP

NegP

NegP VP negation

Low CP + + –

preT + + –

preAnteriorT – +/– +/–

pregenericAsp – –/(+) +

PrevP – – +

Sentence final – +/– 

–

+

–/(+)

–

–

–

with respect to other properties like the possibility to block V to C movement in interrogative clauses, the obligatoriness or optionality (marked here as –/+) of negative concord with n-words, the possibility of occurring in imperative clauses in which the verb has a unique imperative morphology (i.e. it is not ambiguous with any other form), and whether it must, can, or cannot reorder with clitics. Each of the four types of negative elements has a unique combination of + and – with respect to these properties, which attests that they are indeed all different. I have also added the CP and vP negations that Zanuttini does not discuss with their properties. Although Zanuttini does not discuss the etymological origin of the various negative markers, there seems to be a relation between the syntactic position they occupy and their etymological origin. In what follows I will explore the connection between position and the etymological type of negative markers.

.. E 

.................................................................................................................................. Work done by the typologists on negation has shown that sentential negation can be expressed through a number of etymologically very different elements. A survey provided in various works by Devos and van der Auwera () on Bantu, van der Auwera and Vossen (), by Miestamo (), and in particular by Porcellato () provides us with a general overview of the elements which can develop into sentential negation, which can be: (a) a negative auxiliary (see e.g. non-standard English ain’t); (b) a negative copula probably originating from a cleft construction (such as the Finnish negative marker and Jewish Babylonian Aramaic, see Bar-Asher Siegal a); (c) focus markers (e.g. the negative marker identified in Bantu by Devos and van der Auwera ); (d) a verbal or adverbial element originally related to verbs which contain a lexical negation like lack, refuse, leave, or stop; (e) an adverb originally related to the non-animate negative quantifier corresponding to ‘nothing’ (like Germanic nicht, naught, Latin ne-oenum, Welsh ddim, Occitan res); (f ) elements derived from sentential tags (like Afrikaans nie); (g) minimizers (like French and Catalan pas or Rhaetoromance buca or Bantu pa);

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

    



(h) locatives like Bantu ‘there’ (ko in the language Kongo); (i) possessives (e.g. Kanincin Kwend is a possessive pronoun typically used as a focus marker); (j) modality markers (as in Jamul Tiipay, a Yuman language reported in Porcellato  and Sicilian nommu). Negation is even often borrowed from neighboring languages, which is in itself a rather strange fact, since functional elements are generally rather resistant to borrowing. Furthermore, it is impossible that a language did not have a negative marker on its own before the borrowing happened, so one wonders what the reason for such a type of borrowing might be.⁵ Interestingly, a similar type of variation found by typological work is found inside a homogeneous dialectal area like Italy, where we find a subset of the types corresponding to those listed above: . Type b. i.e. a negative copula: This is the case of Sicilian neca, which is literally n-è-ca, i.e. not.is.that and which is still used as a real cleft un è ca in Palermitan, while in other Sicilian dialects it has already turned into a single negative marker (see Cruschina ), . Type c. The Milanese and Trentino sentence final no which has nowadays become the standard sentential negative marker in Milanese corresponds in origin to the pro sentence negative marker. In Trentino it is still used only in contexts in which negation is focused. Although at present we have no prosodic analysis of these dialects, the fact that Milanese no still carries phonological Focus is clearly perceivable when native speakers produce it. Zanuttini () already relates this type of negative marker to Focus and Poletto and Zanuttini () analyze emphatic positive and negative constructions of the type no che non viene (no that he comes, meaning ‘he won’t come indeed’) in standard Italian as cases in which FocusP is occupied by the negative marker no. . Type d. This type of negative marker, related to a verb expressing a negative meaning of the type lack, stop, etc. found in typological work (see Porcellato  for a review of the literature on a sample of over  languages), is represented in the Italian dialects by the negative marker manco, etymologically related to the verb mancare ‘lack’ and to the adjective mancino ‘left-handed’. The form manco is widespread in southern Italian dialects with the value of ‘not even’ but has developed into the standard negative marker in Basilicata dialects like Rionero in Vulture, as shown in Garzonio and Poletto (). . Type e. Rhaetoromance nia and Piedmontese nen, Occitan ren, are etymologically related to the element meaning ‘nothing’, which is nent in Piedmontese and nia in Rhaetoromance. Occitan ren is clearly the counterpart of French rien, but is also the standard negative marker in the Occitan dialects spoken in the Western part of Piedmont. . Type f. The minimizer type of negation is very widespread in the Romance domain. The etymology of French pas and Catalan pas as deriving from Latin passum ‘step’

⁵ Porcellato () reports borrowing of negation in grammars of Austronesian languages like Roglai, Rengao, and Biak, and in Abun, Western Papua.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

is well known. Further minimizer elements are spoken in northern Italian mica, miga, mina, mia, etc. derived from Latin micam ‘crumble’, northern Lombard buca, ca corresponding to Latin buccam, ‘morsel’, and Florentine punto (meaning ‘dot’, see Garzonio and Poletto ).⁶ . Some negative modality marker is found in Sicilian dialects which have a special set of modal particles similar to the ones found in the Balkan Sprachbund (see Damonte  for a description of these modal elements which can be either modal IP particles or complementizers). The form of these elements is a combination of the negative marker and the modal marker which gives rise to forms like nummu (negation nun + modal marker mu with assimilation of the coda of the negative marker /n/ to the onset /m/ of the modal marker, see Damonte ). There is also one further case of conflation of negation and modal markers, the case of pemmu, where the order is modal + negation and is analyzed as a case of modality in the CP. One might ask why some negative etymological types found across the languages of the world are not attested in the Romance domain. This might not be chance, but the effect of independent properties of these languages. For instance, since in Romance the IP is strong and always attracts the inflected verb even in languages with a relatively reduced verbal morphology like French, one would not expect to find negative auxiliaries, or modals or negative tags, since the lexical verb competes for that position. Hence, some negative types require independent properties of the languages where they occur, for instance a weak inflection.⁷ Table . shows the comparison of this structure with the different etymological origins of the negative markers.

Table .. Negative markers and their etymological origin Position

CP negation

Etymology Modal forms

TP negation NegP

Tanterior negation NegP

NegP Aspectual negation

NegP

VP Neg

Clitic forms

Minimizer

n-word ‘nothing’

Focus negation

Negative determiners

Note: Here I use Zanuttini’s terminology in addition to mine to make the parallel to structure () clear.

⁶ Since in Romance the IP is strong and attracts the verb, it is possible that at least some of these types that are related to an independent realization of inflection, like negative auxiliaries, modals, or negative tags, cannot be found because they are incompatible with a strong Infl. ⁷ One further problem concerns the reason why etymological negative markers can have so many etymological sources as shown above. In Poletto () and () I have argued in favor of the idea that there is a link between the amount of etymological variation we find in a certain genetically and grammatically homogeneous area and the amount of semantic and therefore morpho-syntactic features a given functional element is endowed with. In other words, the more an element is complex the more its etymology will vary. The case of negation is rather clear: since it can have so many different etymological origins even within a homogenous domain like the Italian dialects, it must have several features (see Poletto  for a discussion of these features).

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

    



.. R    

.................................................................................................................................. Hence, it is pretty clear that postulating different positions for sentential negation in the clause is supported by rather strong empirical reasons. On the other hand, there might be a way to unify different types into a single one by postulating movement of other elements around the negative marker. In Garzonio and Poletto (a) we proposed that at least some sentence final elements (in particular those related to pro-sentence negation like Milanese no and maybe Brazilian Portuguese nao) are actually located in a left peripheral Focus position, which attracts the whole IP, thus reversing the order of the clause and appearing sentence finally. In this way structures where the negative marker no is sentence final and structures where it is sentence initial are identical modulo the spell-out of one or the other copy of the clause: ()

[HTP [non è arrivato] [ForceP [FocPnoi [PolP noi [TP . . . [ForcePOPi] [Force che [ PolP ei non è arrivato]]]]]]]

In a structure like (), if the higher copy is spelled out, we obtain sentence final no; if the lower copy of the clause is spelled out, we obtain sentence initial no; Notice that in pragmatically marked cases, both copies of the clause can be spelled out, yielding a negation which is sandwiched between the two copies of the clause:⁸ ()

No che non è arrivato. no that not is arrived ‘He did not!’ / ‘Not at all!’

()

Non è arrivato no. not is arrived no ‘He did not!’ / ‘Not at all!’

()

[Non è arrivato], no che [non è arrivato] no that not is arrived not is arrived ‘Of course he hasn’t arrived!’

Notice that the same type of analysis can be used to explain sentence final negations of the question tag types analyzed by Biberauer () for Afrikaans, who brings convincing

⁸ I refer to Garzonio and Poletto (b) for the tests that show that the distribution of sentence initial and sentence final no is the same. We examined compatibility with other left peripheral elements like Focus, Left Dislocation, Scene setting adverbs, other adverbial modifiers and Hanging Topics, and the (im)possibility to be embedded.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

evidence in favor of the idea that the sentence final nie is a CP element as well as sentence final Focus-related negations like Brazilian Portuguese nao. These negative markers are sentence final, but are actually CP elements which attract the whole clause to their left. An alternative analysis, which has to be evaluated on the basis of the distribution of sentence final negation in embedded clauses is to assume that there is no remnant IP but simply vP movement in front of the negative marker located in the low vP peripheral Focus position postulated by Belletti () and subsequent work. The two alternatives make different predictions: the structure in () can only be found in main contexts; the low alternative has no relation to sentence embedding. Which of the two alternatives is valid for the various languages that present this type of negative marker remains to be evaluated on the basis of tests like embedding, since a priori both possibilities are allowed. Although some cases of sentence final negation can be analyzed as a CP negation, it is clearly not possible to reduce all negative markers to a single position with movement of portions of the sentence (either the vP or the I or the CP) as proposed above, because the various types have other properties which differentiate them, as shown in Table .. This means that we are indeed forced to postulate that negation can occur (a) in the CP domain (probably in the lower CP area dedicated to Focus, wh, quantifiers, and more generally operators), (b) immediately above TP (either as a clitic or as an independent morpheme), (c) in the Aspectual area either higher or lower than Tense Anterior (whose specifier is occupied by adverbs like ‘already’ in Cinque’s hierarchy) and (d) in the vP area connected to arguments (see Manzini and Savoia ). The lowest type of negation, that is the one internal to the vP, is not only attested by so-called negative determiners like German kein and Dutch geen⁹ but has been also postulated (see Rowlett  for a discussion) to explain the historical development of French pas, and more generally the fact that minimizers that were originally complex nominal expressions in the object position are reanalyzed as the standard negative marker (see Garzonio ).¹⁰ Garzonio () shows that the Old Italian minimizer mica was already a quantificational element of the direct object, but could also occur in positive environments. He proposes that the grammaticalization path of minimizers goes through a stage in which the minimizer is not (yet) negative, but is already a quantifier of the object. Different types of elements can be related to the CP domain, like types b, c, and f (originating from a copula stemming from a cleft structure, tags which scope over the whole CP, and focus markers of various types). The TP type can be instantiated by negative auxiliaries (like English ain’t and adverbials related to lexically negative verbs). The aspectual type includes elements like minimizers that are located in a low IP area where in general other quantificational elements land (see the position of universal quantifiers in Cinque’s work and Kayne’s  analysis of bare quantifier positions in front of the past participle in French) and quantifier negation of the ‘nothing’ type. The fact that minimizers and quantificational negation occupy two different positions in the aspectual area should not be surprising, since each quantifier most probably has its own position in this area (see again Cinque  for a distinction between universal and negative ⁹ Although these are determiners, the meaning is clearly that of sentential negation, so these elements are to be considered as sentential negation. ¹⁰ The fact that negation can be found in the low argumental portion of the sentence structure is not incompatible with the fact that negation can be realized as a clitic, since clitics generally represent arguments and as such XPs, although they are moved as heads. The same occurs with negation.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

    



quantifiers).¹¹ As for the lowest area, that is the thematic one related to so-called negative determiners, it can be shown that elements preceded by kein- in German non-standard varieties like Bavarian, which tolerate negative concord, have to move out of their canonical object position to an area that is defined in its upper limit by the adverb meaning ‘never’ and in its lower limit by the negative marker itself. The following examples show that elements preceded by kein- need to move to the left of the negative marker nit in Bavarian and this is true for both direct objects, as in () and PPs, as in (), while in general in German varieties non-negative PPs need not climb higher than the negative marker (as shown in ()). ()

a. daß da Hons koa Buach nit glesn hot that the H. no book(acc) not read has ‘H. did not read any book’ b. *daß da Hons nit koa Buach glesn hot that the H. not no book(acc) read has

()

a. daß da Hons auf koan Freind nit gwoat hot for no friend not waited has that the H. ‘H. did not wait for any friend’ b. *daß da Hons nit auf koan Freind gwoat hot c. daß da Hons auf koan Berg nit that the H. on no mountain not ‘H. did not climb on any mountain’

gstiegn is climbed is

d. *daß da Hons nit auf koan Berg gstiegn is ()

a. daß Hans nicht auf den Berg gestiegen ist that H. not on the mountain climbed is ‘H did not climb on the mountain’ b. daß Hans auf that H. on

den Berg nicht the mountain not

gestiegen ist climbed is

If it is true that the lower type of negative markers, that is negative determiners, have to move outside the vP, this means that we can actually reduce the positions of negation to three fields: the CP field where the negative marker is located in the lower operator area, higher than TP in the clitic area, and in the aspectual field in the positions where various types of (bare) quantifiers land when they are raised from their basic position within the vP area. Interestingly, we can see a parallel between the CP and the AspP areas, since in both fields we find dedicated positions to operator-like elements: in the CP it is clear that there

¹¹ As for locative and possessives, I do not have a real answer, but I surmise that some of these elements might also be related to Focus. Devos and van der Auwera  mention that the class of possessives used for negation in Bantu is the same that can signal Focus. I will leave these cases, which do not occur in Italian dialects, to future research.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

are dedicated positions to wh-items (in the Slavic languages where multiple wh-fronting is found there is most probably even more than one position) and in the AspP where quantified arguments land (see again Cinque  on universal quantifiers and Kayne  for French preposing of bare quantifiers to the past participle). The third area where negation occurs is that of argumental clitics, which have no quantificational import, but are still arguments. Therefore, it seems tempting to argue that negation can target positions in various areas of the clause, but all of which are related to specific types of elements that have the same type of features also contained inside the negative marker: whs and focused ones in the CP, clitics in the TP, and quantificational ones in the AspP.¹² ()

[ Topic field [TopicP . . . [Operator field [WhP wh-items [NegP [FocusP . . . [IP field [SubjclP [NegP [ObjclP . . . [TP . . . [Aspect field [NegP [TAnteriorP [NegP [Completive Asp . . . vP]]]]]]]]]]]]]]]]

This observation is interesting, because it can explain why negation is so special within the array of functional elements, which, as noted at the very beginning, have a universal ordering, while negation does not. The fact that quantificational arguments and negation can be located in the same area(s) is not surprising under the view that they share a quantification feature in the relativized minimality typology proposed by Rizzi, where we see that negation shares the Q feature precisely with Focus and wh-items since it gives rise to minimality effects with wh-items and quantifiers (see Rizzi ). If negation has a Q feature as proposed by Rizzi (), then we expect it to target positions in the quantificational subfields of the clausal spine. The low aspectual one identified by Cinque (), where quantifiers like tutto ‘all’, niente ‘nothing’, molto ‘much’ occur is indeed the target of one type of negative marker, precisely those that are etymologically related to quantifiers. The other subfield where quantifiers occur is the low CP area (see Benincà and Poletto  for a proposal on the layering of the CP fields), where wh-items and other types of quantificational items and focused ones can land, and this is also a possible position of the highest negation type. Since negation contains a feature that makes it akin to focused elements and wh-items, it is not surprising that it targets the low CP area. The negative elements that occur in this area are precisely those that are etymologically related to Focus or to complementizers. At this point the picture that emerges is that negative markers target the position their etymological type indicates: those that are related to Focus target the CP subfield where Focus occurs, those that derive from quantifiers (the ‘nothing’ type) target the low IP area where quantifiers are moved.¹³ In other words, they maintain the same syntactic position that their etymological source had before being reinterpreted as a negative marker. The other major point that this section shows is that negative positions are interspersed with others of different types and not concentrated in a single area. Actually, as we have seen, they occur in all subdomains of the clausal

¹² Notice that in the aspectual field there is more than one NegP because the two types of negative markers correspond to different types of quantifiers, since minimizers, located above TanteriorP, are existential, while the lower NegP corresponds to a negative quantifier. See section . for more discussion. ¹³ It is more surprising that negation targets the clitic field, where elements are intrinsically definite and have no quantificational properties. We leave this problem open since it definitely requires semantic considerations that are beyond the scope of the present chapter.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

    



architecture in the same areas where specific types of nominal expressions (wh-, clitic, or quantificational) occur.

.. N    J C

.................................................................................................................................. Evidently, the idea that all domains of the clause can be targeted by negation could also be helpful to understand the well-known Jespersen Cycle, which is probably to be understood as doubling of the positions that constitute the circuit along which negative markers can surface. Until now all the examples of the Jespersen Cycle that we have in the literature (at least to my knowledge) have to do with the loss of a higher negative marker and the realization of a lower one, which is in line with general ideas on economy of movement. This is the case of the French loss of the higher negative marker ne, the case of the loss of en/ne in the Germanic languages, and also the cases found in the Bantu languages which have been analyzed in the typological literature by Devos and van der Auwera (), Devos et al. (), and van der Auwera and Vossen () that share surprising similarities with Romance, and among others, also the fact that the new negator is a lower one, often instantiated either by a minimizer or by a focus-related item, just like in Romance. However, Garzonio and Poletto () have shown that there are indeed cases of change of the negative marker without the doubling stage typical of the Jespersen Cycle, but in these cases there is no change in the position of the negative marker, just the substitution of the old negation with the new one while the two negators maintain the same properties described for TP-related negation in Table .. This is expected if we think that across the clausal spine there are several positions which can host negative markers, although they are in origin per se not NegPs, but generally host different types of quantificational XPs, which have one property/feature in common with negation, as I will propose below. The last point to be made is the following: if the negative circuit exists, we should be able to find cases of doubling of all the negative markers occurring in the various positions. It is indeed possible to find several combinations. We can pretty easily find T-negation and minimizer negation (French, Northern Italian dialects):¹⁴ ()

A n magn meng a la cherna   at  the meat ‘I do not eat meat’

Emilian Carpi MO

The combination between T-negation and quantifier negation is also attested: () Dytaurela n el nia gny Yet  is-he  come ‘He has not come yet’

¹⁴ For Bantu see Devos and van der Auwera ().

Rhaetoromance S. Leonardo di Badia BZ

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

The combination between T-negation and Focus negation is also attested:¹⁵ ()

No se dis cosi no  it says so  ‘We do not say so’

Val di Non TN

The combination of minimizer and quantifier negation is also attested: () Fa pa en Do not not ‘Don’t do that’

sulì that (Zanuttini : )

What is missing are combinations with the complementizer-like negation and the negative article type, which as far as I know are not reported in the literature, so the question whether all doubling possibilities are attested remains open. In section . I will concern myself with the other major question concerning negative positions, namely the possibility of movement.

.. M    

.................................................................................................................................. In this section I reconsider the various types of negative markers and try to determine (a) whether there are possible movements of negative markers from one position to the other across what can be defined as a “negative circuit” and (b) what the reason for these movements is. I will not consider the well-known phenomenon of negation raising found in sentences like ‘I do not think that John will leave’ where the negative marker is located on the main verb, but the interpretation is on the embedded verb (for this see Collins and Postal  and Zeijlstra’s a work) because it involves biclausal structures. I will rather try to ascertain whether it is possible for negative markers to move from one negative position to higher ones of the negative circuit inside the structure of a single clause. Structure () shows that the negative circuit is interspersed with other positions in the clausal spine. However, assuming the latest version of relativized minimality as discussed in Rizzi (), negative markers can in principle move from one position to another inside the circuit without yielding ungrammaticality provided there is no other element located in the movement path that shares features with the negative marker. There is actually empirical evidence that negative markers can move from one position to another. This is already taken for granted in Pollock’s () first analysis of French negation, where the element ne is merged lower in the structure and then is moved to the clitic field where it

¹⁵ This is rather frequent in the Trentino area, although this type of negation is going back to a system where only preverbal negation is found, or focus negation is only used in special contexts, and is known to have existed in Milanese in the sixteenth century (see Vai ), which has nowadays only Focus negation.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

    



adjoins to the left of the inflected verb. The same is true of the so-called postverbal negative marker pas in French, since it can occur either alone or with ne in front of the complementizer in purpose clauses, as illustrated in (): () Pour pas que les pates collent So  that the noodles stick () Pour ne pas que tu m’oublies For   that you me forget Further evidence that minimizer negation can also occur preverbally most probably in the Focus position in the CP¹⁶ where CP-negative markers occur is provided by Italian mica (see Penello and Pescarini  for a discussion of mica preposing): () Mica ti ho detto di uscire  you have told to go.out ‘I did not tell you to go out’ () Non ti ho mica detto di uscire  you have  told to go.out The same is true of other negative markers, which can occur even higher than the position where Zanuttini places them: Manzini and Savoia (: ) notice that quantifier negation, usually located lower than adverbs of the yet/already pair, can be found higher than the adverb in some Piedmontese dialects (cf. Mobercelli, Pamparato, Montaldo).¹⁷ This clearly shows that negative elements can move inside the circuit to higher positions with respect to the one where they are usually located. Even more interesting from the theoretical point of view are those cases in which a negative marker that usually occurs in a certain position surfaces lower than that, because it shows the first merge position of the negator. This is the case for instance in Rhaetoromance of the Müstertal and Florentine (see Garzonio  where minimizer negation occurs lower than the adverb already/yet, while in the vast majority of other dialects it occurs higher; they are represented here by the dialect of Turin). ()

A l’a pa già ciama He .has not already called

()

jau dormel aun bo sleep yet  I ‘I don’t sleep yet’

Turin

Müstertal

¹⁶ Since the preverbal position of mica is incompatible with the presence of non, one might think that the movement of mica is to the TP position of non. Apart from the fact that non is in a clitic position, and mica does not have the typical properties of a clitic, there are other reasons why I think that this is the CP Focus-like position and not the TP one. ¹⁷ Manzini and Savoia do not actually provide any complete examples of this, just a list of dialects which allow for the order: quantifier negation, already.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

() Un ha ancora dormito punto Not has yet slept not

Florence

The same is true of T negation, since it can occur lower than the inflected verb in the position where usually minimizer negation is found (i.e. higher than the adverbial pair already/yet): ()

El ciami mia non anmo Him I.call not not yet

S. Angelo Lodigiano (Manzini and Savoia : )

These facts can only be interpreted in the following way: negative elements which usually occur in the aspectual or clitic fields are actually first merged in a lower position, and the (few) dialects where Manzini and Savoia notice that they are lower are exactly those that display the first merge position of the negative markers. We are thus forced to assume that even clitic and aspectual negators are merged lower, most probably in the vP area, and then raise higher. However, there is no evidence that the highest type of negation, the complementizer-like one located in the CP area, is actually merged lower down in any of the lower positions of the circuit. Although this is definitely the case for the element represented by a negative complementizer, there is evidence that the focus-related negative markers of the pro-sentence negation type (e.g. Brasilian Portuguese nao, Milanese no, Trentino nɔ, etc.) can indeed be found lower down in the structure: see Manzini and Savoia (: –), reported below. () El dorme no semper he sleeps not always ‘He doesn’t always sleep’

(Manzini and Savoia : , )

This does not undermine the analysis presented in section ., but simply means that even the highest type of negation that reaches the CP can be first merged lower in the structure and stop in one of the lower positions in the negative circuit. Manzini and Savoia () propose that negation is actually always an argument and starts out in the object position of the verb. This means that the basic position of all negative markers is that of negative determiners in languages like German, which is located in the object DP but still has scope on the whole clause: ()

Ich habe kein Geld I have no money ‘I do not have money’

In theory we should find examples in different languages that all types of negative markers, including the complementizer-like one, could occur in the object position. Although the hypothesis that negation is merged in the internal argument position straightforwardly explains why negative elements target positions inside argumental fields (like whs, clitics, and quantifiers), this type of evidence that it is indeed always merged in the vP is still lacking at the moment. Furthermore, it remains to be explained (a) why some elements can

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

    



raise higher than others, that is what triggers movement and (b) how complementizer-like negation is to be accounted for. In Poletto () I explored the possibility that negation is made up of a complex set of elements and that the different types of negative markers actually express one feature inside this complex set. For instance, negative elements that are originated from Focus still tend to target Focus and elements derived from the object quantifier still move to the same quantificational position to which quantifiers move. According to this view, the “negative circuit” illustrated in () would be an epiphenomenon due to diachronic inertia: although an element can stand for the whole negative complex set of projections, it still moves to the position it used to move when it did not. This leaves the question open as to why elements like minimizers, focus, etc. can be used to mark negation, and the only possible solution I see is that they must share some quantificational features still to be exactly pinned down through semantics (see Poletto  for a discussion of this).

.. S  - 

.................................................................................................................................. The last question we deal with is whether standard and non-standard negation occupy different positions, or whether there is a unique dedicated position for specific pragmatic types. In the introduction I adopted the definition of standard negation provided by Miestamo () and used in subsequent typological work on negative markers. Nonstandard negative markers are defined as those that are not the standard, that is all those that have an additional pragmatic import in addition to purely negating the clause in which they occur. Several authors have indeed noticed that some negative markers have a special pragmatic import in addition to the negative meaning. This is particularly relevant with respect to the Jespersen Cycle where the “new negator” is often added first in pragmatically marked contexts and only later on is reanalyzed as the standard negator (see van der Auwera b for a discussion on this). At present, I do not think that there exists a detailed cross-linguistic survey of all the possible pragmatic imports non-standard negation can convey. Very often the pragmatics is rather complex and it is not easy to compare between languages. If we consider as non-standard negation all the negative markers that convey any sort of additional pragmatic value with respect to the standard one of negating the clause, we might conceive that pragmatics is directly inserted into some (maybe left peripheral) position. Consider for instance the Italian so-called presuppositional negation first identified by Cinque () for colloquial Italian. The pragmatic import it expresses has to do with a sentence implicature, so the distinction between () and () is not in terms of truth value, the second simply implies that the speaker is not coming, contrary to the expectation of the addressee: () Non vengo Not come ‘I am not coming’

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

()

Non vengo mica Not come not ‘Contrary to what you seem to think, I am not coming’

Zanuttini () notices that standard French also uses pas for the same presuppositional use, and since French pas and Italian mica occupy the same position, proposes that there is a position for presuppositional negation. However, recent work on resumptive sentence final Brazilian Portuguese nao shows that the same presuppositional value (and the same distributional restrictions) apply to an element which is clearly not in the same position; this nao is sentence final and belongs to the etymological type of Focus negation (see above Table .).¹⁸ () A: Voce gustou da palestra da Maria? You like of.the talk of Maria? ‘Did you like Maria’s talk? B: Eu nao fui nao I not went not ‘I didn’t go.’

(Schwenter : )

Therefore, we assume that there exists no direct link between the position of the negative marker in the clause and its status as standard or non-standard negation. Non-standard negative markers are either the “new negative markers” which start out the JC or the old ones, but we also find cases in which the same element can have or not have a non-standard interpretation. However, as noted by one anonymous reviewer, clitic negation is generally not used in cases of non-standard negation. This might be an epiphenomenon due to the fact that clitic negation is the oldest negative marker in this area, and as such it is the standard, but definitely requires further investigation of those dialects which are losing the preverbal negative marker to see whether it is really the case that clitic negation does not lend itself to special interpretations. Furthermore, it is possible to pile up more than one in the clause, if they have a different pragmatic value. So the following sentence in Venetian has three negative markers, the standard one no, the presuppositional one miga (roughly corresponding to colloquial Italian mica), and the focus one no which expresses a Focus on the negative marker itself (which occupies the same position as Brazilian Portuguese nao and has the same etymology just discussed). Since the pragmatics of the two negative markers is compatible, they can be combined. ()

No la go miga magnada NO! Not it have not eaten not ‘I did not eat it’

Venice

The occurrence of several negative markers is also reported by the typological literature (see Porcellato  on this) both with standard and non-standard negative markers.

¹⁸ See also the discussion found in Haegeman and Breitbarth (), who claim that Flemish en only partially overlaps with presuppositional negators but contrary to them (a) it is not restricted to main contexts and (b) it has no negative import whatsoever.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

     ()



ka-zeby-áandi khúumbu ya ŋgúdy-áani kó .-know- .name . .mother-.  ‘he does not know the name of my mother’ (Suundi Hb, Baka : fieldnotes)

We conclude by noticing that pragmatics and syntax do not match each other in such a simple way, at least for where negation is concerned.

.. C 

.................................................................................................................................. In the previous sections I have first shown that negation is special with respect to other functional elements, since it can occur at different heights in the sentence structure even in a syntactically rather homogeneous domain as the Italian dialects. I have shown that the position of different negative markers can occur in the CP, immediately higher than the TP and in the aspectual area. I have also shown that their position is related to their etymological types and reflects the process of grammaticalization they have undergone: for instance, the negative markers derived from a quantifier occur in the same position in the clause in which quantificational elements occur, negative markers derived from Focus elements occur in the same position in the clause in which Focus occurs, and so on. The reason why there is a correspondence between etymology and syntactic position is due to the fact that negative markers maintain the same original position they had when they were not the standard negative marker. I have also shown that there is a minority of cases in which we can have movement inside the negative circuit, that is lower negations can move to higher positions. Since we also find sporadic cases of higher negative elements that occur very low in the sentence structure, I have considered the hypothesis that negation starts out in the internal argument position, which however requires further empirical support. I have also shown that there is no link between the interpretation as standard/non-standard negation and a single position in the clause. We conclude that the observation that negation can occur in various positions, while other functional elements do not, is due to the fact that negative markers maintain the position they had when they were not negative (yet). The conclusion of this chapter could be used to dismiss the variable position of negation as an epiphenomenon due to historical facts. I think that there is more to it, and that negation can occur in so many positions and have so many etymological sources because it has a complex internal structure which at least in part reflects the features of the elements used to become negative markers. I leave the development of this idea to future research.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  ......................................................................................................................

                    Case studies ......................................................................................................................

 

.. I

.................................................................................................................................. T negation of the central proposition of a sentence is usually implemented through the inclusion of some additional item (or items) in the sentence.¹ Where exactly the negative item shows up in the surface sentence form can depend on its category (word, clitic, or affix) and on aspects of the syntax of the particular language. There are also cases where the inclusion of the negative element induces particular effects on the constituent ordering in the sentence structure that are distinct from those observed in the corresponding non-negated sentence. In this chapter I examine a handful of language particular cases in which the presence of sentence negation has apparently non-canonical word order effects. I will also explore the typological/parametric characteristics of the data that I present. In so doing, my goal is to bring out some answers to the following questions: ()

Questions about negation and constituent ordering a. What kinds of non-canonical word orders are attested under negation across languages? b. Where languages have non-canonical ordering under negation, do these languages have comparable non-canonical ordering in other construction types? c. Where different languages have comparable non-canonical ordering under negation, do these languages have other characteristics in common?

¹ In other cases the negative might be a fully-fledged verb. For example, if a language has a verb root denoting ‘not exist’ that is not transparently related to a verb root denoting ‘exist’, then we cannot count the negative verb as an extra element added into the sentence, unless the ‘exist’ sentence is altogether lacking a verb. Similarly, where negation is marked by a change in tone, it may not be apparent that the negated surface sentence includes an increment that is not present in the corresponding positive sentence.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   



In gathering together data from languages that have non-canonical ordering under negation, I will be attempting to provide an answer to (a) so as to arrive at at least a partial classification of the properties of the constructions that exhibit the phenomena crosslinguistically. The question that I raise in (b) is in essence about whether or not negation is unique in the apparently disruptive effects that it may have on sentence structure. The conclusions in those parts of the discussion will thus especially open up more questions and challenges for future research as the data that we will be considering will necessarily be limited with respect to the range of constructions covered. Contributing an answer to question (c) will be the classifications and analyses arrived at in the exploration of answers to questions (a) and (b).

.. C    : D

.................................................................................................................................. In the following sections I explore data from observed cases in which negative sentences and their non-negative counterparts exhibit contrasting constituent ordering. With the notable exception of specific sections in Dryer (c), this particular perspective on negation appears to be relatively novel.² The data that I present has largely been gleaned from reported instances mentioned in more general discussions of negation (Dahl ; Payne ; Kahrel and van den Berg ; Hovdhaugen and Mosel ; Zanuttini ; Cyffer, Ebermann, and Ziegelmeyer ; Miestamo a, b; Dryer c), and I make no claims as to the exhaustiveness of the coverage for known languages of the world. Before embarking on the data presentation, it is necessary to set apart the kinds of cases that will not be included in this treatment of negative/positive constituent ordering

² Miestamo (a: ) characterizes the ‘asymmetric negation’ category included in the WALS atlas (Dryer and Haspelmath ) as involving structural differences between negative constructions and the corresponding affirmative constructions. However, the mismatches that Miestamo identifies include morphological as well as syntactic reorderings. Contrasting auxiliary and main verb ordering in negative/affirmative clauses in Basque (see .. in this chapter) is not singled out because: “Some languages show asymmetries that cannot easily be connected to the marking of any category, e.g. the order of auxiliary and main verb changes in negatives in contemporary standard Basque [ . . . ]; such asymmetries are found in only a few languages and are treated as A/Cat here” (Miestamo b: ). The “A/Cat” label is applied to an asymmetric marking of a verbal category that is distinct from a marking of finiteness or of Realis/Irrealis (Miestamo b: , and for more extensive discussion, see Miestamo ). There are  languages in WALS that fall either into A/Cat or into A/Cat in combination with another characteristic (A/Fin or A/NonReal). However, my focus in this chapter is on the constituent ordering contrasts rather than on cases where distinct affixal ordering is found in affirmative sentences. Dryer (c) provides a detailed and valuable treatment of the placement of negative elements with respect to V, S, and O positioning and includes valuable sections explicitly addressing cases of differences between constituent ordering in affirmative and negative sentences. Most of the case studies discussed in the present chapter coincide with the ‘reordering’ languages treated by Dryer—a total of twenty-eight languages. Undoubtedly, more ‘reordering’ languages will come to light in future research. In the meantime, I attempt to provide as comprehensive a coverage as possible on the basis of the sources that I have indicated in the text.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

differences. One such case is illustrated in the differential placement of the negative pas with respect to the finite and non-finite verb in French, as in: ()

a. Nous n’ allons pas au cinéma. we  go..  to.the cinema3 ‘We aren’t going to the cinema.’

French

b. Nous avons décidé de ne pas aller au cinéma. we have.. decided de   go. to.the cinema ‘We have decided not to go to the cinema.’ Following Pollock (), the ordering difference between pas and all- in (a) and (b) is accounted for as a general distinction in the behavior of finite and non-finite verbs: the finite verb in (a) raises to a higher position over pas, whereas the non-finite verb doesn’t. The inclusion of the negative elements is not here disrupting the canonical ordering patterns of French sentences: they are simply additions. Another kind of case that falls outside the scope of the present discussion is one where the negative sentence includes an auxiliary verb that is not present in the corresponding non-negative sentence. This phenomenon is observed in English do-support and in the following example from Hixkaryana: ()

a. k-amryek-no .-hunt-. ‘I went hunting.’ b. amryek-hra w-ah-ko hunt- -be-. ‘I did not go hunting.’

Hixkaryana

(Miestamo b: )

Cases of this type have additions, but they do not have reorderings of independent words in sentences. Finally, the present discussion focuses on sentential negation, leaving aside cases where the negation applies to a constituent internal to the sentence proposition.⁴ Because variant constituent ordering in a language must be ascribed to syntactic processes, in approaching negative/positive ordering differences, I have chosen to arrange the discussion in terms of the location of the negative element in the sentence: whether it is initial (.), final (.), or clause-internal (.). Within these negative placement divisions, the data will be discussed in terms of canonical word order characterizations for ³ Abbreviations in the glosses conform to The Leipzig Glossing Rules (Comrie et al. ), except for: AFFIRM ‘affirmative’, ASS ‘assertive’, CL ‘clitic’, EMPH ‘emphatic’, FORM ‘formative verb prefix’, IMM ‘immediate’, NONIND ‘non-indicative’, NPFV ‘non-perfective’, P ‘preposition’, PERM ‘permissive’, PERS ‘personal’, PERSIS ‘persistive’, PRT ‘particle’, RECNT ‘recent’, SUB ‘subordinate’, T ‘tense’. Some glosses have been adapted from their sources for the purposes of uniformity within the chapter. ⁴ Although it is necessary to recognize that there can be ambiguities as to whether the whole or a part of a given proposition falls under the scope of the negation. For further discussion see: Dahl (), Payne (), Horn (: –).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   



S(ubject), V(erb), and O(bject). This approach will provide a starting point for the investigation of possible cross-linguistic parameters in the syntax of the phenomena more generally.

.. I 

.................................................................................................................................. The negation of an affirmative proposition p has a semantic representation: NEG(p), but in many languages the sentential negation element is located within the propositional content in the surface form of the sentence. In the case of languages that locate the sentential negation element at the beginning of the sentence (or at the end of the sentence), there is an apparent direct correspondence between the sentence form and its logical representation. The expectation could be that, where a language has an initial sentential negative, the negative is outside of the sentence proposition and would have no effect on the constituent ordering within the main proposition. This, for example, is the case with SV ordering in Mam (Mayan) and with VSO ordering in Welsh: ()

a. ma chin b’eet-a . . walk- ‘I walked.’ b. nti ma chin b’eet-a  . . walk- ‘I didn’t walk.’

()

a. Fe/mi welais i John. / saw  John ‘I saw John.’ b. Ni ddarllenodd Emrys y llyfr.  read Emrys the book ‘Emrys didn’t read the book.’

Mam

(England : ) Welsh

(Roberts : )

But there are nevertheless instances where languages with an initial negation marker have contrasting constituent ordering patterns in affirmative versus negative sentences. In this section, data on such cases is presented and discussed for languages with affirmative VSO ordering (..–..), one language with OSV ordering (..), and a language with SVO ordering (..). In another kind of case, alternative ordering patterns are available in affirmative sentences, but with restricted versions of these in negative sentences (..; ..).

... Polynesian languages In Māori and Tahitian (Eastern Polynesian) negated sentences are introduced by a negative verb (Hohepa ; Chung , ; Lazard and Peltzer ). Māori and Tahitian have

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

VSO constituent ordering, but, in the presence of an initial negative, the subject can precede the lexical verb, as shown in the following pairs of examples: ()

a. I te whakarongo a Hera.  listen  Hera ‘Hera was listening.’ b. Kāhore a Hera i te whakarongo.   Hera  listen ‘Hera was not listening.’

()

Māori

(Bauer : )

te puta. a. ’ua tai’o ’oia i  read   the book ‘He has read the book.’ b. ’aita ’oia i tai’o i te puta.    read  the book ‘He didn’t read the book.’

Tahitian

(Académie Tahitienne : )

Whereas the affirmative sentences in (a) and (a) have the standard T-V-S⁵ ordering, the negative sentences in (b) and (b) have the ordering: NEG-S-T-V. In Māori an alternative NEG-T-V-S ordering is possible, but the NEG-S-T-V ordering is the norm and is obligatory with subject pronouns (Chung : ; Bauer : ). Lazard and Peltzer (: ) state that the subject is preposed in negative sentences in Tahitian. My other source for Tahitian (Académie Tahitienne ) gives no indication that any ordering other than that shown in (b) is possible for Tahitian. The placement of the negative at the beginning of the sentence is consistent with the canonical predicate-first pattern in Polynesian languages. In affirmative sentences lacking a copular verb the initial predicate position is filled by the predicate constituent of the proposition. In corresponding negative constructions, it is then the sentence subject that immediately follows the negative. These effects are shown for nominal and locative predicate sentences in Māori in () and (), and in Tahitian in () and (). ()

a. He tamariki rātou. he children PL ‘They are children.’ b. Ēhara rātou i te tamariki.   i the children ‘They are not children.’

()

Māori

(Bauer : )

a. I te kura ia. at. the school  ‘He was at school.’

⁵ I use the ‘T’ gloss for the initial Tense/Aspect/Modality particle in the Polynesian data.

Māori

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

    b. Kāhore ia i te kura.   at the school ‘He was not at school.’ () a. E faehau terā ta’ata. e soldier that man ‘That man is a soldier.’ b. E’ere/e’ita terā ta’ata i te faehau. / that man i the soldier ‘That man is not a soldier.’

ra. () a. I Ra’iatea vau i te mātāmua  Raiatea  at the former time that ‘I was formerly in Raiatea.’



(Bauer : ) Tahitian

(Lazard and Peltzer : )

Tahitian

b. ‘Aita vau i Ra’iatea i te mātāmua ra.   at Raiatea at the former time that ‘I wasn’t formerly in Raiatea.’ (Lazard and Peltzer : ) The subjects in the negative constructions precede the nominal predicates in (b) and (b) and the locative predicates in (b) and (b). The Polynesian languages have a variety of non-verbal constructions and, for some such constructions, alternative subject-predicate constituent orderings are possible at least under particular pragmatic conditions. I will not however, be concerned here with further details in the make-up of these constructions,⁶ as the present focus is on the ordering contrasts illustrated in ()–(). In order to better understand the negative constructions, we need to take a closer look at how they work in sentences with lexical verbs. Given the higher verb categorization applied to the negative, it could be thought that, unlike Mam (b) and Welsh (b) where the initial negative is a particle, the distinctive S-V ordering in negative sentences in Māori and Tahitian could simply be due to a contrast in the ordering of constituents in subordinate clauses. However, as seen in the following examples, both Māori () and Tahitian () retain the canonical V-S ordering in subordinate complement clauses: () I whakaaro-tia tērā pea ka haere mātou āpōpō. Māori  think- that perhaps  go . tomorrow ‘We thought we might go tomorrow.’ (Ngata : think) () Tē hina’aro nei ’oia ’ia haere mātou i tōna fare. Tahitian  want    go .  his house ‘He wants us to go to his place.’ (Académie Tahitienne : )

⁶ Including that the i particle that appears before the predicate in the (8b) and (10b) equative negative constructions is potentially a case marker (Hohepa : ).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

In effect, following analyses in Hohepa () and Chung (), negative verbs in Māori are viewed as subject raising verbs and the position of the subject in negative sentences is accounted for as an implementation of the raising of the subject from the embedded clause.⁷ In this respect, as shown in Waite (), the syntax associated with negation in Māori conforms to the syntax associated with other subject raising verbs in the language, as in: () I tuku-a te tangata kia haere noa.  go freely  allow- the man ‘The man was allowed to go free.’

Māori (Waite : )

Although the exact nature of the syntactic processes in Tahitian has not been subject to particular study, the comparability of the subject-preposing effects in Māori and Tahitian suggests that subject preposing in negative sentences in Tahitian is also due to a process of subject-to-subject raising. If this analysis is correct, the surface positions of the subjects in the negative constructions, although they are in contrast with the constituent ordering in corresponding positive sentences, they are nevertheless in conformity with the canonical Polynesian #VS ordering pattern. We are not however in a position to conclude that the verbal status of the initial negative necessarily gives rise to an alternative constituent ordering in the sentence in VSO Polynesian languages. Whilst Niuean, a Tongic Polynesian language, has negative sentences with an initial negative verb, negative sentences retain their VSO constituent ordering: ()

a. Ai lā kitia e au e pusi.   see    cat ‘I have not yet seen the cat.’ b. Kua nākai tuai fano a ia.    go   ‘He has not gone.’

Niuean

(Massam : )

Thus, at one end of the Polynesian spectrum, Tahitian has obligatory raising of the subject in negative sentences and, at the other end of the spectrum, Niuean has a complete absence of subject raising. Another Polynesian language in which negation does not induce constituent ordering change is East Futunan. In this non-Eastern Polynesian language, in affirmative sentences pronouns can appear in clitic forms following the T particle and immediately before the verb (a), or in fully case-marked forms after the verb (b). In negated sentences either pronoun ordering is possible (c,d), but non-pronoun subjects always occur after the verb (e) (Moyse-Faurie : ). () a. E kau kai.   eat ‘I am eating.’

East Futunan (Moyse-Faurie : )

⁷ For analyses of a wider range of constructions permitting preposing of subjects in Māori, see Bauer (), Pearce (, ), Douglas ().

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

    b. E kai a au.  eat   ‘I am eating.’



(Moyse-Faurie : )

c. Na le’aise kau ano o mako i nānafi.    go  dance  yesterday ‘No, I didn’t go dancing yesterday.’ (Moyse-Faurie : ) d. E se tio a tātou ki le fatu.   see  .  the stone ‘We do not see the stone.’

(Moyse-Faurie : )

e. Na le’ese māsau a Kalala i le fakatasi.   speak  Kalala  the meeting ‘Kalala didn’t speak during the meeting.’

(Moyse-Faurie : )

Thus, in East Futunan, the constituent ordering in the presence of the negative is nondistinct from that in the affirmative sentence: in both sentence types only clitic subject pronouns may precede the lexical verb. In negated sentences, the clitic pronoun is not further displaced to take up a position between the initial T-marker and the negative and it thus does not have the raising characteristics associated with the subjects in negative sentences in the Eastern Polynesian languages. From the data presented for Tongan in Broschart () and in Otsuka () and for Tokelauan in Vonen (), it appears that affirmative/negative sentences in these non-Eastern Polynesian languages conform to the ordering patterns illustrated for East Futunan. Niuean, East Futunan, Tokelauan, and Tongan are all non-Eastern Polynesian languages and all have ergative case-marking systems.⁸ In the three languages that have raising of subject pronouns in affirmatives as well as in negatives, East Futunan, Tokelauan and Tongan, unlike in the Eastern Polynesian languages, the preverbal clitic pronoun is preceded by the T-marker. The Eastern Polynesian languages, Māori and Tahitian, are non-ergative and they do not allow preverbal subject pronouns in affirmative sentences, but both these languages allow pronominal or non-pronominal preverbal subjects in negative sentences (obligatorily so in Tahitian). In negative sentences, further Eastern Polynesian languages have subject raising options that are not available in affirmative sentences: the pre-T-verb option for pronouns is shown for Hawai’an in Elbert (: ) and for Marquesan in Mutu (: , ). A somewhat intermediate case is Pukapukan (nonEastern Polynesian). Pukapukan has both ergative and nominative/accusative casemarking options (Chung , ) and, like the Eastern Polynesian languages, it allows ⁸ My thanks to a reviewer who suggests that the absence of subject raising under negation in the languages with ergative case marking could be accounted for as a restriction against the raising of an inherently case-marked subject (as per discussion of inherent/lexical/structural ergative case-marking such as in Woolford ; Rezac, Albizu, and Etxepare ; Polinsky ). There is much that remains to be explored with respect to the syntax of ergative and non-ergative Polynesian languages and it may well turn out that ergative clauses in Polynesian languages are frozen for certain extraction processes. However, whilst it might be supposed that any potential subject raising in these languages would apply to nominative/absolutive arguments, rather than to ergative case-marked arguments, Woolford (: ) cites a Tongan example showing the raising of an ergative argument under the verb lava ‘be possible’ or ‘be able, manage’.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

raising of both pronoun and non-pronoun subjects in negated sentences, but not in affirmatives (Chung ). Interestingly (see fn. ), Chung (: –, –) shows complete acceptability for raising of the nominative argument in both active and passive constructions, but a lesser acceptability for raising of the absolutive argument under negation. In summary, the Polynesian languages that have been discussed here can be characterized generally as having initial negative verbs, but alternations in constituent ordering under negation are restricted to a subset of these languages: those which have canonical nominative/accusative rather than ergative case-marking and which do not admit of the T-pronoun-V ordering in affirmative sentences.⁹ The alternate ordering possibility under negation in Māori and in Pukapukan is accounted for in Chung () as subject-to-subject raising in the syntax of these languages. It seems plausible that an analysis of this type would be applicable to the other Polynesian languages that have the reordering of the subject under negation.

... Eastern Nilotic Sentential negation in the Dorik dialect of Lopit, a VSO Eastern Nilotic language spoken in South Sudan, uses an initial auxiliary verb (Moodie ). In the presence of the negative auxiliary verb, the subject appears before the lexical verb: ()

a. é-ìlá xɔ´tɔ´ɲɪ´ `ɪ ŋεˆ -wash. mother. baby. ‘The mother washes the baby.’

(Moodie : )

b. ´-ɲà ɪ náŋ l-á-wú à tòrɪ´t ´ɪ -not.be . --go to Torit ‘I am not going to Torit.’

(Moodie : )

c. ´ɪ -ɲà xáɟàŋáʔ l-è-fánù ´ɪ -not.be flies. --go.PL ‘The flies did not come.’

(Moodie : )

Lopit

In (b,c) the Lopit negative ɲà occurs with an unspecified person agreement prefix ´ɪ -. This form of the negative is found also in existential uses: () ´-ɲà ɪ xɪ ̀fjóŋ ´ɪ -not.be water. ‘There is no water.’

Lopit (Moodie : )

However, there are two constructions in which the negative auxiliary bears a standard person agreement prefix: (i) when the negative auxiliary ɲà is marked for irrealis, and

⁹ There are known to be twenty-eight distinct Polynesian languages (Clark ) and any claims that I make about Polynesian languages are restricted to those that I have mentioned specifically in the present discussion.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   



(ii) with the persistive negative marker ɲeí. The examples in () show the contrasting uses of the prefix in the non-irrealis and the irrealis forms of ɲà, and those in () show third person agreement with the negative ɲeí (in contrast with ɲà in (c)). () a. *a-ɲà l-á-wú à tòrít náŋ -not.be . --go to Torit ‘I am not going to Torit.’ b. á-ɲaɪˆ-ɲà náŋ l-á-mwei . . . --not.be . --be.sick ‘Were I not sick . . . ’ ()

è-ɲeɪˆ xáɟàŋáʔ é-lè-fánù -not.be. flies. --go. ‘The flies have stopped coming.’ (lit.: ‘the flies are still not coming’)

Lopit (Moodie : ) (Moodie : ) Lopit

(Moodie : )

Subject preposing applies under sentential negation regardless of the presence or absence of full subject agreement. Especially given the presence of agreement marking on the verb under negation, whether or not the constructions should be understood as involving a classic form of subject-to-subject raising is unclear. Displacement of the subject applies under sentential negation with an initial negative verb also in the Eastern Nilotic languages Otuho and Ateso (Moodie ). The negative auxiliary and the lexical verb in Otuho bear the standard person agreement prefix, as in: () a-beng dwo ni a-lo -not.be .  -go ‘I didn’t go there.’

Otuho (Moodie : )

Among further Eastern Nilotic languages, in Maa and Turkana the sentential negative marker is a prefix, and in Bari it is a particle (Moodie ). I lack information on whether or not Maa and/or Bari allow subject preposing under negation. However, in Turkana affirmative sentences, both VSO and VOS constituent ordering is attested (Dimmendaal : ), but there is no preposing of the subject in the presence of a negative prefix: ()

a. ɲ-ὲ-ɲamà a-yɔŋ’ a-kɪ̀-riŋ not--eat- me meat () ‘I have not eaten the meat.’ b. à-ɲam’ a-yɔŋ’ a-kɪ ̀-riŋ -eat  () meat10 ‘I have eaten the meat.’

Turkana

(Dimmendaal : )

In the data that we have seen in this and section .., two characteristics that are common to the Polynesian and Eastern Nilotic languages that have preposing of subjects in negative ¹⁰ I here retain the distinct glosses given with these examples in Dimmendaal ().

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

sentences are: (i) the languages are verb-initial; and (ii) the negative element is a verb. The Nilotic languages differ from the Polynesian languages in that they have agreement marking both on the negative verb and the lexical verb. If in both language types the subject preposing counts as subject raising, the presence of the double agreement marking in the Nilotic languages suggests a non-classic form of subject raising in these languages. The preposing of the subject to a position after the negative verb keeps the verb-initial surface pattern, however, in both cases.

... Non-Nilotic African VSO languages Gude (Chadic) and two Surmic languages, Tennet and Majang, are three further languages with VSO constituent ordering and with subject preposing to a position after an initial negative word (Dryer c: ). Jonathan Moodie (personal communication) notes that Arensen () reports that Murle, another Surmic language closely related to Tennet, also has nominative preverbal subjects following an initial negative. The VSO/NEG-SVO alternations for Gude and Tennet are illustrated in () and (). ()

()

a. kə kii Musa faara  throw Musa stone ‘Musa threw a stone.’ b. pooshi Musa kii faara  Musa throw stone ‘Musa did not throw a stone.’

Gude

(Dryer : )

anná lokúli balwáz a. k-ɪ´-cɪ´n-ɪ .--see-. . Lokul yesterday ‘I saw Lokuli yesterday.’ b. ŋanní anná k-ɪ´-cɪ´n  . .--see ‘I did not see Lokuli yesterday.’ c. k-á-cɪ´n-ɪ anná .--see-. . ‘I see Lokuli now.’

Tennet

lokúli balwáz Lokul yesterday lokúli íyókó nέkɔˆ Lokul now 

d. ɪrɔ´ŋ anná k-a-cɪ´n-ɪ lokúli íyókó nέkɔˆ  . .--see-. Lokul now  ‘I do not see Lokuli now.’ (Creissels et al. : ) The Tennet negative words are distinct in (b) and (d), but in both cases the preposing of the subject derives NEG-SVO ordering, as also in Gude (a,b), contrasting in both cases with the affirmative VSO ordering. Whereas, in the Polynesian and the Eastern Nilotic languages, the subject preposing applies where the initial negative is a verb, in Gude and in Tennet the initial negative is not described as having the characteristics of a verb. However, Randal (: ) draws on the fact that the negated verb following ŋanní occurs in the subjunctive form to suggest that ŋanní would have earlier been a finite negative verb taking a subjunctive complement.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   



... Nadëb Nadëb, a Tucanoan language of Brazil, has basic OSV constituent ordering with SVO as an alternative (Weir : ). It has three distinct sentence-negating morphemes (dooh, na-, and manih), each occurring in distinct types of constructions. The negative word dooh occurs sentence-initially giving rise to alternate constituent ordering in habitual/iterative constructions. In a transitive sentence, following dooh, the constituent ordering may be OSV or SVO. Weir (: ) states that the SVO ordering is more common in dooh clauses in texts and that it is her impression that the SVO ordering is more neutral than the OSV ordering in sentences negated with dooh. The examples in () show an affirmative sentence in (a), and the two alternatives with the dooh negative: OSV in (b) and SVO in (c). () a. Awad kalapéé ha-púh. jaguar child -see. ‘The child sees the jaguar.’

Nadëb

bú. b. Dooh awad kalapéé ha-púúh  jaguar child -see.  ‘The child doesn’t see the jaguar.’ bú awad. c. Dooh kalapéé ha-púúh  child -see.  jaguar ‘The child doesn’t see the jaguar.’

(Weir : –)

Weir (: ) describes the content of the clause following the negative dooh as a nonfinite nominalized clause (the non-indicative form of the verb is required in non-finite clauses). The particle bú, which is glossed as an ablative, is said to have a subordinating function (Weir : , fn. ). This analysis suggests that the preferred constituent ordering differences in affirmative sentences versus in negative sentences could at least in part be attributable to contrasting syntax in finite clauses versus in non-finite clauses. The negative manih is used following the verb in negative imperatives. Weir gives examples of transitive permissive imperatives where manih appears in both an affirmative and a negative subordinate clause with SVO constituent ordering: () a. Na a txaah i-ug  . son -drink. ‘Let your son drink the medicine!’ b. Na a txaah i-ug  . son -drink. ‘Let your son not drink the medicine!’

ta biin.  medicine

Nadëb

manih ta biin.   medicine (Weir : )

In this case, there is no apparent contrast in ordering between the affirmative in (a) and the negative in (b). It is possible that the SVO ordering in (a,b) is a function of the fact that the S, V, and O constituents are in non-matrix clauses (as is the case with the ‘neutral’ construction with the dooh negative in (c)). The third type of negative marker in Nadëb is a verb prefix na-. In one construction using na-, described in Weir () as a Grammaticalized Negative Relative Construction

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

(GNRC), the negative is inside a nominalized clause that is the predicate in an equative sentence. There are two kinds of such GNRCs: in the form in (b) the nominalized predicate precedes the subject of the sentence; in the form in (c) the verb is initial immediately preceding the subject and the object is marked as a dative. The affirmative in (a) has the preferred main clause OSV ordering. ()

a. Tóóh dab kad a-wu. wild.pig meat uncle -eat. ‘Uncle is eating wild pig meat.’ b. Tóóh dab na-wuuh kad. wild.pig meat -eat. uncle ‘Uncle isn’t eating wild pig meat.’ (lit.: ‘Uncle is a non-wild pig meat-eater.’) kad tóóh dab hã. c. Na-wuuh -eat- uncle wild.pig meat  ‘Uncle isn’t eating wild pig meat.’ (lit.: ‘Uncle is a non-eater with respect to wild pig meat.’)

Nadëb

(Weir : , )

Superficially, in (b) the subject kad ‘uncle’ follows the predicate which has an internal OV ordering; and in (b) the subject follows the verb of the predicate and the object of the predicate comes after the subject. Weir (: ) states that there is no direct affirmative counterpart to (c) and that there is no apparent meaning difference between (b) and (c). The predicate-initial ordering of (b) patterns with other predicate-initial equative constructions in Nadëb: ()

a. Ta txaah iih.  son  ‘I am his son.’ b. Ta txaah na-do iih.  son -be.  ‘I am not his son.’ bú. c. Dooh ta txaah iih a-do   son  -be.  ‘I am not his son.’

Nadëb

(Weir : )

With these further examples of equative constructions, we see that the predicatesubject ordering in the (b) negative is in conformity with that in the affirmative in (a). Both of the negative constructions, with na- in (b) and with dooh in (c), are wellbehaved with respect to the expected predicate-subject ordering in equatives. Leaving aside the special case of the construction in (c), we remain with the affirmative/doohnegative ordering contrast in (a) versus (c). Given the nominalized clause analysis applied to the dooh construction, the negative sentence in (c) can also be viewed as having the (standard) predicate-argument form with dooh as the predicate of the sentence. The contrasting SVO ordering within the subordinate clause could then be

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   



understood as due to a preferred constituent ordering difference between matrix and subordinate clauses.¹¹

... Languages with alternative affirmative constituent orderings Mettouchi (: ) reports that, across the different dialects/languages of Berber, negative sentences have surface orderings as S-NEG-VO or as NEG-VSO. Abstracting away from the presence of the negative, the alternate orderings available to negative sentences correspond to those in affirmatives. However, there is an affirmative/negative ordering contrast involving clitic pronouns: with initial negative wər or one of its variants, clitic pronouns precede the verb, whereas in corresponding affirmatives the clitic pronouns follow the verb: ()

a. y-əfka yas t iD .-give. . ..  ‘He gave it to her/him.’ b. ur s t iD y-əfki  . ..  .-give. ‘He did not give it to her/him.’ c. ur t id i-wwit ara  ..  .-hit. . ‘He didn’t hit him (contrary to expectations).’

Berber12 (Mettouchi : )

(Mettouchi : )

(Mettouchi : )

In the negatives in (b,c) the lexical verbs are in the negative perfective form, whereas the lexical verb in the affirmative in (a) is in the perfect(ive).¹³ As Mettouchi (: ) notes, two other negative words, ma and attha, do not co-occur with the negative perfective and neither of these forms is found with preverbal clitics. This is seen in the following example with attha as the negative element and where the proximate clitic follows the verb: ()

attha usa-nt D  arrive.-.  ‘They haven’t arrived (didn’t arrive)’

Eastern Taqbaylit Berber (Mettouchi : )

Constructions with the wər negative thus have two characteristics that are distinct from constructions with the other negatives: the non-canonical location for clitics and the use of the negative perfective verb form. An intriguing question is how these two characteristics are related syntactically in the grammar of Berber.¹⁴

¹¹ I note that it could be that object preposing from an underlying SVO constituent ordering base is more readily available in matrix clauses than in embedded clauses. ¹² Specifically, Western Taqbaylit, Kabyle (Amina Mettouchi p.c.). ¹³ Lafkioui (: ) states that the negative perfective is more common than the negative imperfective. ¹⁴ For more extensive discussions of the syntax, see Ouhalla () and Ouali (), for example.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

Jenner (: –) describes distinct constituent ordering patterns in Cornish in terms of inflected versus impersonal verb forms and pronominal versus non-pronominal objects. George () shows SVO ordering for the affirmative data in (a) and (a) and preferred VSO ordering with an initial negative in (b) and (b). ()

a. An del a wel an gath. the man  see the cat ‘The man sees the cat.’ b. Ny wel an den an gath.  see the man the cat ‘The man does not see the cat.’

()

Cornish

(George : )

a. My a wel an gath.   see the cat ‘I see the cat. b. Ny welav an gath.  see. the cat ‘I do not see the cat.’

(George : )

Although other orderings are available for all of (b) and (a,b),¹⁵ the ordering in the affirmative (a) is shown as the unique ordering in a database for sentences of this constituency. Dryer (c: –) refers to two further languages, Masakin and Barambu, both NigerCongo, with an initial negative and with affirmative SVO~VSO constituent ordering. In Masakin there is bipartite initial and final negation marking co-occurring with SVO constituent ordering. Dryer suggests that the SVO ordering, both in the affirmative and the negative, aligns with an imperfective aspect interpretation. In Barambu, on the other hand, the negative occurs in the VSO ordering pattern either with an initial and a final negative marker or just with the initial negative marker.

... Wanyi Wanyi (Australian, non-Pama-Nyungan) presents another case in which clitics are differently located relative to the verb in affirmative and negative sentences. Wanyi is a predicate-initial ‘free word order’ language (Laughren, Pensalfini, and Mylne ). In the affirmative (a) the verb is initial and is followed by the clitics. In the negative (b) the negative is initial, the clitics come next and are then followed by the verb. (The symbol ‘=’ shows clitic attachment in the Wanyi data.)

¹⁵ An alternative sentence corresponding to (32a) is: (i)

a. Y hwelav vy an gath.  see.  the cat ‘I see the cat.’

[Stephen Morey p.c.]

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

    ()



a. Daba=bula=ngaa=n kirriya-wiya-a. Wanyi hit=.=.=n woman-pair- ‘Two women hit me.’ (Laughren, Pensalfini, and Mylne : ) b. Bundangku=yala=ninya ngurubu nang-kanyi, daba=yalu-ngka =.=. tell this- hit=- ‘They didn’t tell you about this, (that) they had a fight.’ (Laughren, Pensalfini, and Mylne : )

Unlike the clitics in Berber (see () versus (a,b,c)), clitics in Wanyi are obligatorily in second position (except in the case of sentences with an initial topic). Whilst pragmatically unmarked affirmative sentences are verb-initial, in sentences with a focused constituent, the focused constituent is initial and is immediately followed by the clitics: () a. Winyjika-a=ninya windijbi nang-kanyi mama-anyi? who==. give this- bread- ‘Who gave you (of ) this bread?’

Wanyi

b. Nana-ngkani ngandaara-a=nga=n windijbi nang-kanyi that- old.woman-=.=n give this- mama-anyi. food- ‘That old woman gave me (of ) this food.’ (Laughren, Pensalfini, and Mylne : ) In the analysis that Laughren, Pensalfini, and Mylne () present, the initial constituents are phrases in the CP specifier agreeing with a feature in C. The non-co-occurring C features are [+focus], [+negative], and [+wh].¹⁶ A verb (or predicate) in the initial position (Spec,CP) has information focus and agrees with the C [+focus] feature. Thus, although clitics are differently ordered with respect to the verb in affirmative versus negative sentences, the position that they occupy is in conformity with the featural realizations in the syntax of the language. Laughren, Pensalfini, and Mylne () show that the analysis that they present for Wanyi can be extended to Warlpiri and, potentially, to further verb-initial Australian languages.¹⁷,¹⁸

¹⁶ Wh-constituents do not co-occur with the negative in Wanyi (Laughren, Pensalfini, and Mylne : –). ¹⁷ A difference in the case of constructions in Warlpiri is that, in the terms of Laughren, Pensalfini, and Mylne’s () analysis, the Warlpiri auxiliary verb can check the [+focus] feature in C. ¹⁸ Miestamo (: ) points to another case involving a construction-specific affirmative/negative clitic repositioning in Andoke (a language of Colombia). Here, also, with a second-position condition on clitic placement, the presence of an initial negative results in clitics preceding rather than following the verb: (i) a. o-do-ʌ m-əi -know- - ‘They know me.’

o-do-ʌ b. hʌ´má m-ə  - -know- ‘They don’t know me.’

Andoke (Miestamo : )

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

... Bafut Bafut (Niger-Kordofanian) has SVO constituent ordering in affirmative sentences, but, with a discontinuous bipartite negative surrounding the subject, the ordering is SOV: ()

a. Bó tù’ù mə̀ ŋkì. they carry . water ‘They have just fetched water.’ ŋkɪ̀ tū’ū. b. Kāā bó sɪ̀  they  water carry ‘They have not fetched water.’

Bafut

(Chumbow and Tamanji : )

The switch to a verb-final construction in negatives is seen also in existential constructions: ()

a. Sùù tsì ghū. Suh be there ‘Suh is there.’ b. Kāā Sùù sɪ ̀ ghú tsī.  Suh  there be ‘Suh is not there.’

Bafut

(Chumbow and Tamanji : )

It is not clear if the contrasting constituent orderings would be accounted for as different verb and/or XP displacements in affirmative versus in negative clauses. Present also in () are alternations in tone assignment producing final mid-tone on the final verbs in (b) and (b), tū’ū and tsī, (versus tù’ù in (a) and tsì in (a)), as well as with final ghū in (a) versus non-final ghú in (b).¹⁹

... Overview An initial negative word can potentially be located within the IP, the basic clause; within the left periphery of the clause; or within a higher clause taking the proposition being negated as a complement subordinate clause. In this section, I began by looking at instances in which the negative is considered to be a higher verb taking a clause as its complement. With sets of VSO Polynesian (..) and Eastern Nilotic languages (..), in some of these languages, the presence of the initial negative verb induces the preposing of the subject. In the Polynesian case, the subject preposing has been analyzed as an implementation of subject-to-subject raising. The presence of the higher negative verb, however, does not

¹⁹ Intriguingly, Dahl () cites another Niger-Congo language that has an affirmative/negative constituent ordering switch, but with the negation marked by a tone change on the object: (i)

a. mà tὲ màna ‘I bought bananas.’ b. mà maná tὲ ‘I didn’t buy bananas.’

Kwaa

(Dahl : )

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   



necessarily entail subject-to-subject raising, an option that seems to be restricted to languages of the Eastern Polynesian type. Where Eastern Nilotic languages have an initial negative particle or affix (instead of a verb), the data did not provide evidence for subject (or other) displacement in the negative constructions. In the case of the VSO Surmic languages (..) the initial negatives may not synchronically have the morphological characteristics of verbs, but it is possible that the subject preposing observed in negative constructions in these languages is indicative of earlier functions of the negatives as verbs (as suggested in Randal : ; and Dryer c: ). In the further verb-initial language, Nadëb (..), we saw that both affirmative and negative sentences can occur with OSV or SVO constituent ordering. There is, however, a preference for OSV ordering in affirmatives and for SVO ordering in negatives. It is proposed that the proposition that is being negated has the structure of a subordinate clause and the affirmative/negative OSV versus SVO preferences pattern with matrix OSV and subordinate SVO preferences. The next section (..) covered two further cases of languages with alternating constituent ordering patterns in affirmative sentences. In a dialect of Berber there are negative constructions, however, that impose a different ordering on clitics from that seen in affirmative sentences. Cornish, on the other hand, was shown to be less restrictive in its ordering patterns in negative sentences than in at least one type of affirmative sentence. Clitic positioning was also shown as distinct in negative versus positive sentences in the Australian language Wanyi (..). In the analysis of Laughren, Pensalfini, and Mylne () a negative or a focused constituent occupies the sentence-initial position and the clitic pronouns are obligatorily located after the initial constituent. Because the verb in an affirmative sentence can occur in the initial position, in such cases, the clitics follow the verb. Thus when the sentence has an initial negative the clitics follow the negative and they necessarily precede the verb. The data from Berber indicated that affirmatives do not have a second-position requirement for clitics. The second-position placement for clitics was observed in constructions with only one out of three of the initial negative markers. A Niger-Kordofian language, Bafut (..) provided us with an instance of constituent ordering alternations in an SVO language with a bipartite discontinuous negative, the first part of which is sentence-initial. In the Bafut case the negative constructions are verb-final: thus, NEG-S-NEG-OV. Only two of the languages examined here do not have verb-initial ordering as a basic pattern in affirmative sentences: Nadëb (OSV~SVO) and Bafut (SVO).

.. F 

..................................................................................................................................

... SOV languages Final position for negative marking is relatively common in languages with SOV constituent ordering. Out of a total of  consistent SOV languages with a single negative marker, Dryer (c: –) shows a total of  (.%) languages with the negative as either the final constituent or as an affix on the final constituent (the verb). Affixal negation is often

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

found as part of the verb morphology in verb-final languages, but without impacting on the constituent ordering of the clause. Among well-known cases of this type are Japanese and Turkish. Descriptions of further languages in this category are: Peeke (): Waorani; Nedyalkov (): Evenki; Hartzler (): Sentani (Papuan); Sandonato (): Zazaki (Indo-Iranian); Barnes (): Tuyuca (Tucanoan).

... Other constituent orderings Except for a double negation variant in Kayabí, (Tupi) (Dryer c: ), I have not found any instances of languages with consistent SOV affirmative ordering and some other ordering pattern co-occurring with a final negative. All instances that I have found of languages with a final negation marker and with contrasting affirmative and negative ordering have SVO as the sole affirmative ordering or they have affirmative SVO ordering alternating with SOV. In the former category are two Surmic languages, Me’en and Tirmanga, and two Niger-Congo languages, Mbosi (Dryer c: ) and Leggbó (Dryer c: ). In the latter category Dryer (c: ) includes four Moru-Ma’di languages (Nilo-Saharan) and Dongo (Niger-Congo). Interestingly, in the case of the four languages that only have SVO in the affirmative, the constituent ordering with the final negative is SOV, whereas in the case of the five languages with alternating SVO~SOV affirmatives, the constituent ordering with a final negative is SVO. The former pattern is exemplified with data from Me’en in () and the latter with data from Ma’di in (). ()

a. εdε or kobu-o they see chicken- ‘They see the chickens.’ b. εdε kobu-o or-on they chicken- see- ‘They don’t see the chickens.’

()

a. ká gbándà `ɲā.  cassava ..eat ‘He is eating cassava.’

Me’en

(Dryer : ) Ma’di

b. ɔ´-ɲā gbándà rá .-eat cassava  ‘He ate cassava.’ c. m’-āwí dʒótī kʊ -open door .. ‘I am not opening the door.’

(Dryer c: )

The Me’en data simply shows a word ordering contrast for the two sentence types. In the Ma’di data the negated sentence in (c) has the same SVO constituent ordering as the positive sentence including the affirmative marker in (b). These two sentence types are in contrast with the SOV ordering found in the positive not including the affirmative marker in (a).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   



Although Logbara (Moru-Ma’di) also has its final negative marker in complementary distribution with an affirmative marker, in the Logbara case the ordering in the two sentence types is distinct. ()

a. Drùsɪ ̈̀ mâ zâ ɲaa rá. tomorrow  meat eat  ‘Tomorrow I shall eat meat.’ b. Drùsɪ ̀̈ á ɲaa zâ kö. tomorrow  eat meat  ‘Tomorrow I shall not eat meat.’

Logbara

(Payne : )

Payne (: ) also notes other differences in the affirmative versus the negative sentences in (): the use of a short form of the subject pronoun in the negative construction and a contrast in the tone marking on the object in the affirmative/negative forms. Following the analysis of Miestamo (: –), these characteristics, along with the word order contrast, distinguish what is the incompletive construction in (a) from the completive construction in (b). There could potentially be a comparable aspectual contrast in the patterns for the other Moru-Ma’di languages of this group.

... Summary Dryer (c: –) identifies a total of  SVO languages having negative sentences with a marker in final position (either a single negative marker or one member of a double marking pair). The very few cases that we have seen of languages with a contrasting constituent ordering under final negation, with one exception, involved SVO ordering at least as a possible ordering in affirmatives. The two kinds of patterns that we have observed, SVO~SOV-NEG and SVO/ SOV~SVO-NEG, were found in just two language families, the former in Niger-Congo and the latter in Moru-Ma’di (Nilo-Saharan), along with Dongo (Niger-Congo) located in close proximity with the Moru-Ma’di languages (Dryer : ). These cases are especially intriguing as, overall, the presence of final negation does not appear to be particularly conducive to alternations in constituent ordering.

.. C- 

..................................................................................................................................

... Languages with SVO affirmatives As was seen to be the strong tendency for languages with contrasting affirmative/negative constituent ordering where the negative marker is final (.), where the negative marker is internal (non-initial or non-final), I have found contrasting constituent ordering only in languages that in the affirmative have either SVO as the only ordering or SVO and SOV as alternative orderings.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

There are six Niger-Congo languages that I have identified as having affirmative SVO and negative S-NEG-O-V ordering. In the Mbili and the Kru data in () and () respectively there is a simple reordering of constituents in the presence of the negative: () a. a gʊa atɨ ? fell tree ‘He fells a tree.’ b. a ka atɨ gʊa ?  tree fell ‘He does not fell a tree.’ ()

a. ɔ´ t è ᷈ kɔ̀ he buy rice ‘He bought rice.’

b. ɔ´ sé kɔ̀ t è ᷈ he  rice buy ‘He didn’t buy rice.’

Mbili

(Dryer : ) Kru20

(Hyman : )

In both Mbili and Kru the negative appears as a particle-like word. Apart from the word order difference and the presence/absence of the negative, there are no other apparent differences in the forms of the affirmative and negative sentences in either language. In another Kru language, Dida, the negative is represented as a suffix and, apart from what seems a morpho-phonological effect of the suffixation, once again, the only differences between the affirmative and the negative sentences are the switch in constituent ordering and the presence/absence of the negative marker. () a. dāāɡɔ̄u lìpì flàásʊ̄ Dago speak. French ‘Dago spoke French.’ b. dāāɡɔ̄-ɔ´ flàásʊ̄ lìpì Dago- French speak. ‘Dago did not speak French.’

Dida

(Creissels et al. : )

In data from Pana (Gur) the negative bears the aspect marking that is carried by the verb in the affirmative (see the verb forms in (a) versus (b)). According to the gloss in (c), the lexical verb in the negative is uninflected. The negative construction also has a final focus particle that is not present (although it can be) in the data for the affirmative.

²⁰ Hyman () does not identify the particular Kru language.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

    ()

a. à ɲùé sìmá  drink. beer ‘I drink millet beer.’

 Pana

b. à ɲɔ´ sìmá  drink. beer ‘I drank millet beer.’ c. à wè sìmá ɲɔ´ yà  . beer drink  ‘I did not drink millet beer.’

(Beyer : )

Whereas the negative is expressed by an auxiliary verb in the construction in (c), in the imperfective, it has the form of a suffix on the verb. In this case the constituent ordering of the affirmative is preserved: ()

à ɲùée-ré sìmá yà  drink.- beer  ‘I don’t drink millet beer.’

Pana (Beyer : )

If my reading of the data is correct (in particular, that the verb ɲɔ´ is uninflected in (c)), it is possible that the constituent ordering in (b) versus (c) would be accounted for as involving verb raising to inflection in (b) versus absence of verb raising in (c). A similar analysis of the contrasts in Mbili and Kru data in () and () could be operative, but this seems less likely for the Dida data in () where the verb carries aspectual marking both in the affirmative and the negative. Two further Niger-Congo languages with contrasting SVO and S-NEG-OV ordering are Klao and Grebo (Dryer c: ). The Surmic language, Mursi, also has SVO affirmatives and SOV ordering in the negatives, but in the case of Mursi the negative marker is positioned between the object and the verb, as SO-NEG-V (Dryer c: ).

... Basque In Basque, nominal expressions are case-marked and the case roles of the arguments are indexed on the inflected verb. The neutral constituent ordering is SOV (Saltarelli : ; Ortiz de Urbina : ), but a variety of surface orderings are possible under different pragmatic conditions. In affirmative sentences, however, there is strict adjacency between the lexical verb and a following auxiliary: () a. Bilbo-in Euskara hiru urte-z ikasi zuen. Basque Bilboa-in Basque. three year-for learn  ‘He studied Basque in Bilbao for three years.’ b. Bilbon hiru urtes Euskara ikasi zuen. c. Bilbon ikasi zuen hiru urtes Euskara. d. Hiru urtez Euskara ikasi zuen Bilbon. e. Ikasi zuen Bilbon Euskara hiru urtez. (Ortiz de Urbina : )

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

In a negative main clause,²¹ instead of coming after the lexical verb, the auxiliary comes after the negative particle ez, as in (a). The particle ez also has the function of the simple denial ‘No’ (Etxepare : ). When there is no auxiliary verb, the inflected lexical verb follows the negative particle, as in (b). () a. Jon ez da etxe-ra etorri-ko. Jon   home-to come- ‘Jon won’t come home.’ b. Jon ez dator etxe-ra. Jon.  comes home-to ‘Jon does not come home.’

Basque

(Ortiz de Urbina : )

As in affirmative sentences, a variety of constituent orderings are possible in negative clauses, but the adjacency between the negative particle and the inflected verb is preserved in negative clauses. The examples in () show some possible orderings. ()

a. Mirenek ez dio Joni etxea Miren.  ... Jon. house. eman. give ‘Miren did not give Jon the house.’ b. Ez dio etxea Mirenek Joni eman. c. Etxea Joni ez dio Mirenek eman. d. Ez dio eman Joni Mirenek etxea.

Basque

(Laka : )

Laka () shows that, when a negative sentence is compared with a corresponding emphatic affirmative, there is no contrast in the constituent ordering in the two constructions: ()

a. Jon ez da etorri. Jon   arrive ‘Jon hasn’t arrived.’ b. Jon ba da etorri. Jon so  arrive ‘Jon has so arrived.’ (emphatic ‘so’)

Basque

(Laka : )

In (b) ba is an emphatic particle, a reduced form of the affirmative dai ‘Yes’ (Laka : , n.). Thus, as well as the complementary distribution between ez and da in (a)/(b), there is a further direct correspondence in the uses of the two particles. Laka () develops an account of the syntactic derivations of the constructions in which the location of the affirmative/negative particle is attributed to the presence of a functional P category called ‘ ’ marking the truth value of a proposition. In Laka’s account of the derivations of negative sentences, the inflected verb raises by head-movement from a right-

²¹ For extensive discussions covering negative constructions in non-matrix clauses in Basque, see especially Ortiz de Urbina () and Etxepare ().

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   



P branch position and adjoins to the head. In an antisymmetrical approach (Kayne ) to the hierarchical structure, Haddican () also adopts the use of a high functional projection, a ‘Polarity’ phrase. In the terms of his analysis, the Spec, PolarityP provides the target position for the negative marker and for other phrases when no negative is present.

... Overview Accounts like those of Laka () and Haddican () are inspired by the notion that a proposal that is made for Basque, if correct, should be found to be applicable in other languages as well. A challenge for the understanding of word order permutations in other languages with sentence-internal negation, such as those discussed in section .., is the extent to which they may or may not turn out to have syntactic processes comparable to those applying in Basque.

.. C

.................................................................................................................................. The data that has been examined in this chapter can in no way be said to cover the full gamut of possible forms of negative constructions cross-linguistically.²² But we have nevertheless been able to observe some patterns in the syntactic characteristics of negated sentences. The largest number of occurrences of different constituent orderings in affirmative versus in negative sentences, and across the greatest number of language families, are attested in cases where the negative is the initial item in the sentence. Within this category, the majority of cases occurred in languages with VSO in the affirmative and with SVO in the negative. In many of these cases the initial negative was a verb and the NEG-SVO ordering obtained as an effect of subject raising. Dryer (c: ) has proposed that the preponderance of instances of affirmative/ negative constituent ordering alternations in VSO languages with negative verbs can be understood as a function of the differing placement of the verb relative to the subject in VSO versus SVO or SOV languages. That (canonical) negative verb placement in SVO or SOV languages should not be disruptive of constituent ordering is because the placement of the negative would have no (visible) impact on the subject position, with the outcomes to be expected as: S-NEGV-VO and SOV-NEGV respectively, where VO and OV are the complement parts of the canonical ordering patterns. In the corresponding VSO case, when the initial negative is a verb, outputs with NEGV-SVO can be considered as instances of V-S-Complement, “a variant of the normal VSO order.” A possible problem with this account is that it seems to make the assumption that the S in negated sentences is an argument of the negative verb. However, another way of looking at this scenario could be to say that, even if the S is not an argument of the negative verb, the negative verb still ²² Much of the data sourced is from sections of Dryer and Haspelmath () which covers a total of , languages (Comrie et al. : ).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

has a subject position to be filled (at least in languages that require overt subjects). From this kind of perspective, the VSO/NEG-SVO effects are due to the special behavior of subjects. In the more limited number of cases (both by language family and by location) with either a final or a clause-internal negative, there was a preponderance of languages with SVO affirmatives or with affirmative SVO alternating with SOV ordering. For some of the languages discussed, a significant factor cutting across the constituent ordering divisions was the role of focus marking in the syntax of the sentence. Such effects were seen for auxiliary placement in Basque and for clitic placement in Wanyi. The role of the complementary distribution of focus markers and negative markers in a number of the African languages discussed is suggestive of further avenues of cross-linguistic inquiry with respect to constituent ordering alternations. Of similar potential for further investigation are cases where contrasting affirmative/negative constituent ordering aligns with aspectual distinctions in the sentence structure.

A

.................................................................................................................................. My thanks to the handbook reviewers for their valuable input, to Stephen Morey and Rachel Nordlinger for helpful discussion and to Jonathan Moodie for generously giving me access to his unpublished work and also for informative discussions.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

  ......................................................................................................................

       ......................................................................................................................

 

.. I

.................................................................................................................................. B now it seems unquestionable that sign languages are fully-fledged natural languages and must be part of the empirical evidence linguists work with. However, it is also true that the amount of knowledge about them accumulated to date is still limited when compared to spoken languages, both in the number of languages studied and in the depth of description and analysis. Fortunately, negation is one of the domains that has received systematic attention in sign language research from the start as one of the core grammatical properties not only from a descriptive point of view, but also in theoretical work. There is even work carried out from a typological perspective (Zeshan , a, b, ), which is quite exceptional given the limitations in the quantity of properly described sign languages. When compared to negation systems in spoken languages, the property that sticks out most prominently is the role that non-manual markers play in its expression. Non-manual marking has been shown to play a crucial role in different aspects of the grammar and in the lexicon of languages in the visual-manual modality. We find this type of marker functioning at all grammatical levels, from phonology to pragmatics (for an overview, see Pfau and Quer ). As we will see in this chapter, some of these markers stem from the grammaticalization of co-speech gestures, and when recycled by the particular grammar of a sign language, they are clearly regulated by its rules. Despite some surface similarities across a significant number of sign languages at the manual and non-manual levels, detailed accounts of concrete systems of negation have shown important variation among them, as expected. The most important patterns of variation are summarized in this chapter. The chapter discusses manual signs of negation, both regular and irregular, in section .. Next, it characterizes the form and behavior of non-manual markers in section .. In section . the interaction between both types of negative marking is then analyzed, leading to a brief overview of syntactic analyses of negative structures. Section . tackles

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

the issue of negative concord in sign languages. The interplay of negation with other categories as well as the phenomenon of negative doubling are discussed next in section .. Finally, section . covers negation at the lexical level.

.. N  

.................................................................................................................................. All sign languages documented so far have one or more manual signs that encode negation. Some of them realize clause negation only, while others combine with other functional categories or have specialized uses. Probably the most widespread clausal negator is the one consisting in the index handshape with the palm oriented outwards, moving slightly from side to side, as illustrated in Figure . for Jordanian Sign Language (LIU)¹ (Hendriks : ) and exemplified in ().²

 ..  in LIU

(From Hendriks 2007; reprinted with permission from De Gruyter Mouton).

¹ The acronyms of sign languages are taken from the name in the ambient spoken language, or else in English. ² The usual glossing conventions in the sign language literature are followed, according to which manual signs are represented by the capitalized word corresponding to the gloss of the sign (the most general translation into a spoken language word). The scope of non-manual markings is represented with a line that spreads over the manual material with which it is coarticulated. The relevant abbreviations for the purposes of this chapter are the following ones: ## (verb agreeing with subject and object; the number before the verb refers to the grammatical person of the former and the one after the verb refers to the latter); a (locative index pointing to locus a); # (pronominal index; the number corresponds to 1st, 2nd, or 3rd person); ++ (reduplication of the sign); cond (conditional); hs (headshake); nbp (non-neutral brow

       ()      //  ‘My father and mother aren’t Deaf, they speak.’

 LIU

However, the regular negator in American Sign Language (ASL) has a very different shape, as depicted in Figure .:  is realized as a closed fist with the thumb extended, and it moves from under the chin forward. It appears in a sentence like () (Aarons : ).

 ..  in ASL (Material courtesy of Dr Bill Vicars and www.lifeprint. com)

()

neg     ‘John does not buy tomatoes.’

ASL

In some cases, clausal negation blends with some other functional category, thus giving rise to a more specialized meaning. An example of a sign that merges negation and aspect is  (-) in Catalan Sign Language (LSC), which encodes the negation of the perfect. This can be observed in () (Quer and Boldú ).

position); neg (negative marking); nfe (negative facial expression); re (raised eyebrows). Examples taken from other sources have been adapted to a unified annotation system.

 ()

  cond       , 12 ‘If you haven’t finished the text tonight, I’ll help you.’

LSC

Some clausal negators have also been reported to convey an additional layer of pragmatic meaning, such as presuppositions. For instance, the Turkish Sign Language (TİD) sign  is reported to contradict an assumption on the part of the addressee, be it overtly expressed or implicit (Zeshan c: ). This is exemplified in (). ()

hs ++  - ‘(I) don’t beat my children.’

TİD

An additional component often accompanying clausal negation is emphasis. Apart from its possible instantiation at the non-manual level, some lexical negators are specialized for such emphatic meaning, as in the case of the LSC  (-) meaning ‘not at all’, exemplified in () and illustrated in Figure .. ()

hs  ++    ‘The students did not prepare any topic at all.’

 ..  in LSC (From Quer & Boldú 2006; © LSC Lab)

LSC

      



 .. ^ in LSC

(From Quer & Boldú 2006; © LSC Lab)

From a formal point of view, apart from these morphologically simplex (or regular) negatives, sign languages also systematically display a number of semilexical and semifunctional signs that incorporate negation in a more or less opaque way. Zeshan () labeled them irregular negatives. These signs tend to belong to certain semantic classes of predicates of cognition, emotion or volition, modals, and existence/possession. In LSC, for example, we find cases of quite transparent cliticization of the negator  onto some predicates such as  (see Figure .), where the concatenated signs are recognizable as the independent signs  and , as well as instances of suppletive negative forms that do not bear any formal relationship to the non-negative counterpart. A good example of this is the pair  and  in LSC (Figure .). Negative affixation, either sequential or simultaneous, gives rise to irregular negatives as well. These negative morphemes, albeit identifiable, possess limited productivity, as they only combine with a restricted set of lexical items. Take for instance . in Israeli Sign Language (IsSL) illustrated in Figure ., where an outward twisting movement is combined with the basic sign  (touching the forehead with a B-hand) (Meir : ), thus leading to simultaneous affixation. Negative sequential affixation is attested for a number of verbs in ASL: , which is formationally related to the sign , behaves as a sequential suffix that always occurs after its stem, is added to a limited group of verbs (, , , , , , , , , , , ), and with some of them it results in idiosyncratic meanings (Aronoff, Meir, and Sandler : ff.). Its suffixal status is confirmed by the fact that it only combines with one-handed non-agreeing verbs. Figure . illustrates the transparent case of ^, which means ‘not see at all’.



 

 ..  vs.  in LSC; © LSC Lab

 .. . in IsSL

(From Meir 2004; reproduced with permission from John Benjamins).

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

      



 .. ^ (‘not see at all’) in ASL

(From Aronoff et al. (2005: 329); reprinted with permission from Language, © Diane Lillo-Martin).

.. N - 

.................................................................................................................................. Non-manual components have been shown to play a decisive role in the encoding of negation in sign languages. The specific markers may vary, but they mostly reduce to head movements and facial expressions. They also vary in the ways they interact with manual markers and the ability to convey sentential negation independently or not. However, their importance in the expression of negation systems must be highlighted as one of the most remarkable modality-specific properties of sign languages. Negative non-manuals, like many manual negators, have a clear origin in the gestures used in the ambient societies where the sign languages are embedded. In the case of negative non-manual markers, a gestural source has arguably evolved into a functional element without going through the stage of lexicalization (Wilcox , ). Coincidence in surface form between gesture and grammatical marker does not mean that they function in the same way. As discussed below, it is clear that negative non-manuals, as grammatical markers, are constrained by grammatical rules, and in production they have a discrete onset and offset, are constant, and display a linguistically determined scope (Baker-Shenk ; Wilbur ). Moreover, evidence from acquisition and processing of sign language supports the claim that grammatical non-manuals show distinct patterns when compared to affective communicative gestures (Reilly and Anderson ; Corina, Bellugi, and Reilly ; Atkinson et al. ). The generalizations reached about the use of non-manual markers in the expression of negation are normally based on elicited and grammaticality judgment data. Now that

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

corpus data of spontaneous or semi-spontaneous discourse are available for some languages, evidence has been brought to the fore that attests more variation in naturalistic discourse than described in previous studies. In one case, for instance, the results are argued to support the existing analyses (Oomen and Pfau  on Sign Language of the Netherlands (NGT)), but in another case it has led to questioning the grammaticalized status of headshake as a grammatical marker in Australian Sign Language (Auslan) and to equating it to its co-speech occurrences in spoken discourse (Johnston ). This is not just an empirical issue, but also a methodological and analytical one that requires further detailed research. Among head movements, the most widespread negative marker is headshake, that is a side-to-side movement of the head. It is virtually found in all negation systems described so far (Zeshan : ) and is normally coarticulated with the negative manual sign, if present, and it can spread over other constituents of the clause. Nevertheless, some cases of independent use of headshake have been described, as in Chinese Sign Language (CSL): as illustrated in example (), the non-manual marker of negation must occur independently of the predicate it negates (Yang and Fischer : ). Manual negation is the other grammatical means to encode negation in the language. ()

hs) hs (*  ‘I don’t understand.’

CSL

The default case, though, is that negative headshake co-occurs with manual signs, most prominently with those encoding negation. As will be discussed below, negative nonmanuals, like other markers of the same type, can spread over syntactic domains depending on the language and its grammar constraints. A second, less common type of negative head movement reported in the literature is head turn, which can be interpreted as a reduced form of headshake. It has been documented in British Sign Language (BSL), CSL, Greek Sign Language (GSL), Irish Sign Language (IrSL), LIU, Quebec Sign Language (LSQ), Russian Sign Language (RSL), and Flemish Sign Language (VGT) (Zeshan b: ). A third type of head movement used to encode negation is backward head tilt, occurring in some sign languages of the Eastern Mediterranean such as GSL, Lebanese Sign Language (LIL), LIU, and TİD. The source of this non-manual marker must be found in the negative gesture found in the ambient societies to which the relevant sign languages also belong. In GSL, head tilt typically coappears with a manual negator like - (a), but it can also extend over a whole clause (b), maintaining the position of the head backward after the initial tilt until the final negative sign (Antzakas : ). ()

ht a.     - ‘I told him not to go to work.’ ht b.    - ‘I don’t want to go (there) again.’

GSL

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

      



A second group of non-manual markers of negation is formed by facial expressions. The most widespread ones are frowning, squinted eyes, nose wrinkling, and lips spread, pursed, or with the corners down, while more language-particular ones or even sign-specific ones are puffed cheeks, air puff, tongue protruding, and other mouth gestures. An example of negative facial expression is documented for Brazilian Sign Language (Libras), and it consists in lowered corners of the mouth or O-like mouth gesture. This marking coexists with headshake, but Arrotéia () argues that negative facial expression (nfe) is the obligatory marker in this language, irrespective of the presence of the manual negator, whereas headshake cannot fulfill this function on its own, as illustrated by the contrast () (Arrotéia : ). ()

nfe a. 1 1a Ãa 1 () ‘I didn’t see João.’ hs b. *1 1a Ãa 1 ()

Libras

For TİD, Gökgöz () has argued that non-neutral brow position (either brow-lowering or brow-raising) is the crucial non-manual marker in this language, and it can co-occur with head tilt or headshake (cf. ()). It can spread over syntactic domains marking the scope of negation. ()

hs nbp 3pl 1   ‘I didn’t say these things.’

TİD

The status of the different negative facial expressions is language-particular, as we will see in section .. One of the most interesting aspects to examine in signed negation systems is how manual and non-manual elements interact, as sign languages clearly differ in this respect. Section . delves into this issue.

.. I    -   

.................................................................................................................................. Sign languages have been shown to encode negation with both manual and non-manual components, but they can be split into two main groups as to which one is obligatory: there are languages which require the presence of a manual negative sign, while others can do without it and encode sentential negation with non-manual marking exclusively. These two types are known as manual dominant and non-manual dominant in the expression of negation, respectively (Zeshan b: ). Hong Kong Sign Language (HKSL) instantiates the case of a manual-dominant system, since the presence of a manual negator is

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

compulsory, as the contrast in () shows (Tang : , ): the use of headshake only does not yield a grammatical negative sentence (b). ()

hs a. 3    ‘It’s not true that he is flying tomorrow.’

HKSL

hs b. *  ‘The house isn’t far.’ By contrast, other languages like ASL, German Sign Language (DGS), LSC or NGT can mark sentential negation only non-manually. LSC, for instance, licenses both negative readings with and without the presence of the sign . See the equivalent LSC examples in (a) and (b): in the latter the only overt negative marker is headshake (Quer /). ()

( ( )) hs a.     ( ) hs b.    ‘Santi doesn’t eat meat.’

LSC

An important aspect of non-manual dominant languages is that the non-manual can spread beyond the negative sign or the predicate sign it negates, according to languagespecific rules. In manual-dominant languages this can vary, and the non-manual can remain circumscribed to the sign it is coarticulated with or spread, as in TİD (cf. ()). The explanation for this split in non-manual behavior has been that the non-manual marker coappearing with the manual negator in manual-dominant languages is lexically specified: non-manuals have been shown to play distinct roles at all linguistic levels, and in this particular case it would amount to the phonological specification of the sign for a specific non-manual like headshake or head tilt at the lexical level, just like any other phonological parameter such as handshape, location, or movement. By contrast, in non-manual dominant languages, given the spreading behavior of the relevant markers, they are interpreted as the morpho-syntactic encoding of negation, thus regulated structurally.³ The comparison of three non-manual dominant-languages such as ASL, DGS, and LSC (Pfau and Quer ) makes the point evident. In the three cases, headshake must be obligatorily realized in negative sentences, as the paradigm in () shows. ()

a. *    ‘John is not buying a house.’

ASL

b. *    ‘Santi doesn’t eat meat.’

LSC

³ See Oomen, Pfau, and Aboh () for a proposal for a prosodic rather than a syntactic account of headshake spreading in NGT.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

       c. *    ‘Mother is not buying a flower.’

 DGS

The distribution of the non-manual, though, is not identical in the three languages at hand: while in ASL and LSC headshake can be associated with the manual negative only, this is not possible in DGS, as illustrated in (). For the DGS sentence to become grammatical, the headshake must extend at least over the predicate as well. ()

hs a.     ‘John is not buying a house.’

ASL

hs b.     ‘Santi doesn’t eat meat.’

LSC

hs c. *    ‘Mother is not buying a flower.’

DGS

On the other hand, when the manual negator is absent, headshake can appear only on the predicate, both in LSC and DGS, but not in ASL, as shown in (). ()

hs a. *   ‘John is not buying a house.’

ASL

hs b.    ‘Santi doesn’t eat meat.’

LSC

hs c.    ‘Mother is not buying a flower.’

DGS

For the ASL sentence corresponding to (a) to become grammatical, headshake must spread over the whole verb phrase, as in (a). In LSC and DGS this is also possible, although optional. In the case of LSC, this leads to a contrastive interpretation of the object, in a corrective context, as illustrated in () (Quer /: ). ()

hs a.    ‘John is not buying a house.’

ASL

( ) hs b.    ‘Santi doesn’t eat meat.’

LSC

( ) hs c.    ‘Mother is not buying a flower.’

DGS

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

()

hs hn   ,  ‘Santi doesn’t eat vegetables, but fruit (he does).’

LSC

Pfau and Quer () account for the differences in non-manual behavior among the three languages on syntactic grounds. A basic ingredient of the explanation lies in interpreting headshake as a morpho-syntactic feature which is present in the syntactic structure of the clause. For ASL, an SVO language, Neidle et al. () proposed that the manual sign  and the feature [+neg] are both sitting in the head of NegP. When  is present, headshake is coarticulated with it (a). However, if manual negation is absent, headshake cannot be associated with manual material in Negº (the verb is assumed to be unable to move upwards) (b), so it is forced to spread over its whole c-command domain, which in this case is the whole VP (c). () a. 

hs [NegP [Neg ]

b. * [NegP [Neg c. 

[NegP [Neg

[VP  ] ] hs +neg] [VP  ] ] hs +neg] [VP  ] ]

ASL

Pfau and Quer () build on this type of account but make the further assumption for LSC and DGS that [+neg] is actually a featural affix in the sense of Akinlabi (). Both languages are head-final. In LSC, when  is present, [+neg] gets affixed to it. When there is no manual material in Negº, the verb is argued to move up to Negº, thus satisfying the affixal property of [+neg]. Putting details aside, the representation of LSC negative sentences with and without  would be as in (). () a.  [NegP [VP  ] b.  [NegP [VP  tV][Neg

hs [Neg ] ] hs [V ] ] ]

LSC

The same derivation in (b) accounts for the parallel example in DGS (c), but not for the unavailability of the counterpart of (a) in DGS (c). The contrast is argued to derive from the different syntactic status of  in DGS: unlike in LSC, the manual negator in DGS has phrasal status and is therefore not located in Negº but in SpecNegP. Still, the affixal [+neg] in Negº attaches to the verbal head that has moved to it, thus combining headshake with the predicate. The headshake marking coarticulated with  is lexical in nature and it blends with the one on the predicate. The comparison of the interplay between manual and non-manual marking across three non-manual dominant languages shows the importance of delving into the details of the grammar of negation in specific languages. As more studies are carried out from such a perspective, it becomes clearer that the variation across sign languages is in some respects of the same nature as the one observed in spoken languages, and in other respects, modality-particular.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

      



.. N 

.................................................................................................................................. The possibility of having more than one negative element in a clause expressing a single sentential negation has received extensive attention in the literature on spoken languages, which either do allow for it in systematically distinct ways or do not. The question obviously arises whether the same phenomenon occurs in languages in the visual-manual modality, but needs to be posed in a nuanced way, given the different ways to convey negation available. Since negation is realized by both manual and non-manual components across sign languages, it seems almost trivial to conclude that negative concord is attested quite broadly, in particular in non-manual dominant systems, where the basic negator is the non-manual component which can stand alone and a manual negation can appear optionally. Languages like ASL, DGS, and LSC discussed in section . above instantiate this case and can be argued to realize this type of modality-specific negative concord. Under this view of negative concord Pfau () classifies DGS as a strict negative concord language in Zeijlstra’s (a) sense, whereby negative words/signs always co-occur with the clausal negator. However, the most common way to look at negative concord in sign languages implies the co-occurrence of manual signs in the same clause that yields a single negation reading. The phenomenon as such has received limited attention in the literature so far, but some studies address its existence (Arrotéia  on Libras; Hendriks  on LIU; Oomen and Pfau  on NGT; Pfau and Quer ,  and Quer / on LSC; Wood  and Fischer  on ASL; Kimmelman  on RSL, among some other scattered references), exemplified here in () for LSC, where the clausal negator  appears together with the negative temporal adverbial  under a single negation reading. In () a further example of two negative signs in TİD with a concord reading can be observed (Gökgöz : ). Note that an emphatic negator like ‘at-all’ in () is often attested in the negative concord examples. ()

()

hs hs     ‘I have never smoked.’     - ‘I didn’t know (how to) sign at all.’

LSC

TİD

It is important to keep in mind that the co-occurrence of negative signs in negative concord structures does not imply that those signs can be interchanged. This is expected if, as in LSC, the two elements have a different projection status and occupy different syntactic positions. Therefore, (), the minimally differing counterpart to the LSC example () where the two negatives have switched position, yields an ungrammatical result. ()

hs hs *    (Intended: ‘I have never smoked.’)

LSC

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

On the other hand, sign languages like Italian Sign Language (LIS) (a manual-dominant language) have been shown to be non-negative concord languages at the manual level, as can be seen in example () (Geraci : ). ()

a. *    b. *    (Intended: ‘Gianni never arrived.’)

LIS

Note that from this view of negative concord, DGS (a non-manual dominant language) is a non-negative concord language, since it does not allow the occurrence of two negative manual signs, as illustrated in () (Pfau and Quer : ). ()

hs hs hs *     ‘Roland never drinks beer.’

DGS

The expectation is that the appearance of two manual negators in this type of language should result in double negation readings, which are in principle available but turn out to be harder to process and thus not immediately acceptable out of context, as the unacceptability of () suggests. However, there are data reported that seem to point in this direction, as the LIS example in () seems to indicate (Geraci : ).⁴ ()

??     ‘Gianni cannot not sign the contract (Gianni must sign the contract).’

LIS

From this brief overview it can be concluded that: (i) sign languages instantiate a modality-specific type of negative concord, namely between the manual and non-manual components appearing in the same sentence; (ii) at the same time, they display the basic variation in negative concord at the manual level, constrained by the type of properties found in spoken languages. More research in this domain will help enrich the existing knowledge about this phenomenon.

.. S    

.................................................................................................................................. In this section, a brief discussion is offered of syntactic phenomena that hinge on the presence of negation, namely the interaction with other categories like agreement or tense and the existence of doubling negative structures.

⁴ For the discussion of the possibility that LIS is a Double Negation language, see Pfau ().

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

      



... Negation and other categories As a functional element, negation interacts with other elements in the functional domain. A striking case of the impact of negation on word order as a consequence of the presence or absence of overt agreement is reported in Quadros () for Libras: whereas negation is preverbal with verbs realizing overt agreement like  (a), it must be postverbal with so-called plain verbs, which do not inflect for agreement (b). The difference is accounted for by the blocking effect of the negative head, which bars upward movement of V to the agreement head in the case of plain verbs (agreement verbs do not need to undergo such head movement). (c), with clause-final negation, is the only possibility. ()

hs a.  a  ab  ‘John does not give the book to her/him.’

Libras

hs b. * a    (‘John does not like the car.’) hs c.  a    ‘John does not like the car.’ A different type of interaction of negation with other categories worth mentioning are the combinatorial restrictions for specific lexical categories. IsSL is a case at hand: different negators (, -(/), and -) can only combine with a category type (adjective, noun, and verb, respectively), as exemplified in () (Meir : –). ()

a.  a  /*-/*-(/) ‘The chair is/was not comfortable.’

IsSL

b. 1  -(/)/*-/* ‘I don’t have a computer.’ c. 3  -/*-(/) ‘He didn’t sleep at all.’ A further interaction has been documented in Georgian Sign Language (GESL) between irregular negative verbs and tense: while in the present tense these irregular negatives exclude the co-occurrence of the clausal negator , they require it in the past, so it seems that manual negative concord is restricted to past tense (Makharoblidze and Pfau ). ()

a. *  -   ‘Today my brother doesn’t want to swim.’ b.  - *()   ‘Yesterday I did not want to paint it,’

GESL

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

Additional and more detailed research about such restrictions is needed, also in other sign languages, in order to better understand the details of the interactions of negation with functional and lexical categories. Up to this point, this overview has mostly focused on the characterization of elements that function as sentence negators and fall into the category of negative particles or adverbials. However, sentential negation can be encoded by other categories such as negative determiners or quantifiers, glossed as , , or  . The LIS example in () illustrates the case of a negative subject that does not occur in its canonical subject position (the language is SOV), but in the (right) Specifier of NegP, occupying the typical position of negative elements in this language and triggering sentential negation. ()

hs    ‘Nobody signed the contract.’

LIS

ASL is a language where two distinct negative determiners ( and º) have been characterized. They are exemplified in () (Wood : ). ()    /º ‘John did not break any (part of the) fan.’

ASL

It should be mentioned in passing that negative or positive polarity items are absent from the descriptions available. Abner and Wilbur (: ) write that there are no such items in ASL. However, Schlenker () mentions the existence of a bona fide negative polarity item in ASL, namely . At this point the question is whether further research will be able to uncover such elements, or rather they are extremely rare in sign languages, which would be an interesting fact. Negative adverbials, which are rather common, have appeared earlier on, as in example (). The temporal negative is quite widespread. Interestingly, in ASL  receives different interpretations depending on the position it occupies: if preverbal (a), it expresses the negation of the perfect, but postverbally it conveys a negative modal meaning (b) (Wood ). ()

a.     ‘Bob has never eaten fish.’

ASL

b.     ‘Bob won’t eat fish.’ It should be kept in mind that the type of negation examined so far circumscribes itself to clausal negation (or standard negation), that is, negation of declarative sentences. Nevertheless, we also find negators specialized for other sentence types like the imperative. In LSC, the sign that goes with a negative imperative like () is illustrated in Figure . (Quer and Boldú ). ()

   ’! ‘Don’t touch the magazine!’

      



 .. Negative imperative ’! in LSC

(From Quer & Boldú 2006; © LSC Lab)

... Doubling in negative structures A striking feature related to the syntactic encoding of negation in sign languages is the doubling of negative markers attested in several sign languages⁵ (e.g., ASL, Petronio ; Auslan, Johnston and Schembri ; CSL, Yang and Fischer ; Libras, Quadros ; New Zealand Sign Language (NZSL), McKee ). Doubling is a phenomenon that occurs in sign languages not only with negation, but also with other categories such as modals, whwords, quantifiers, some lexical verbs, and adverbials. In some cases, it is the clause negator which occurs twice, as in example () from CSL (Yang and Fischer : ). ()

nfe nfe /  - / ‘There is nothing to show that you master the whole shape first.’

CSL

In some other cases, the doubled element can be a combination of negation and another element like a modal. Such a case has been reported for ASL (Petronio : ):

⁵ Note that with doubling here we refer to the use of the same negator in two different positions of the clause. This is a different case from those where different negators appear in the same sentence instantiating negative concord. ASL is thus a non-negative concord language that displays (negative) doubling.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

()

hs  ’  ’ (*) ‘Ann CAN’T read.’

ASL

In the examples reported, one of the doubled elements appears in the right periphery of the clause, whereas the other instance occurs clause-internally. Doubling structures are interpreted as encoding emphasis or emphatic focus. This property has been understood as key to the analysis of doubling. Quadros (), for instance, argues that in Libras the final occurrence is actually base generated in Focº and its TP complement is moved to SpecFocP, resulting in the surface linear order observed. Lillo-Martin and Quadros () and Nunes and Quadros () offer a very similar explanation for ASL and Libras, with the only difference that the doubled element located in Focº is actually moved from inside TP before TP moves to SpecTopP (on top of E(mphatic)-FocP). From this perspective, doubling crucially relies on the pronunciation of the same element in two different positions in the structure (for the technical details, see Nunes and Quadros )⁶ and it should be distinguished from another possible, tag-like structure where the doubled element is preceded by a slight pause (Petronio and Lillo-Martin ; Neidle et al. ). It is interesting to note that in doubling structures the accompanying non-manual negative marker can spread between the two doubled elements. In CSL spreading seems not to be obligatory (cf. ()), but possible, as in () (Yang and Fischer : ). ()

nfe   -   - ‘Don’t pay attention to a detail at the beginning.’

CSL

This phenomenon has been identified as perseveration of non-manual markers (Neidle et al. ): the non-manual specified for the doubled elements spreads among the signs that intervene between them in a continuous fashion.

.. N    

.................................................................................................................................. Next to its role at clausal level, negation also plays a role in the formation of words that have a negative component but are not negative themselves and cannot reverse the polarity of a sentence. The morphological processes by which such items are created are essentially the same ones that were discussed in section . above. Here we will illustrate two cases of affixation, one sequential and one simultaneous. Note that, given their lack of sentential scope, these items with negative morphology are not coarticulated with negative non-manuals. As is also typical for morphological derivation, the resulting item may have an idiosyncratic meaning which is not transparently derivable from the composition of its morphemes.

⁶ These accounts take it for granted that the doubled elements are heads and not phrases. For a competing account, see Neidle et al. ().

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi

      



Meir (: ff.) offers an example of negative sequential derivation in IsSL with the sequential suffix -, which stems from the negative existential predicate (). It attaches to nouns and adjectives and gives adjectives as a result, some of them with not totally transparent meaning (e.g. +- in (d) does not mean ‘unenthusiastic’, but rather ‘not caring about something, indifferent’). () a. b. c. d.

+- +- +- +-

‘unimportant, insignificant’ ‘insecure’ ‘unsuccessful’ ‘doesn’t care about it, indifferent’

IsSL

A different case of negative affixation is defined by the presence of pinkie extension in the negative item in a positive-negative pair in East Asian languages like CSL and HKSL. HKSL displays many such instances (Tang : ). The members of pair / and / are only distinguished by the final handshape: thumb-up in the former and extended pinkie in the latter. This opposition in pairs of antonyms is used in CSL such as those in () (Yang and Fischer : ). ()

a. b. c. d.

/   

   

CSL

The same contrast can be found in simultaneous affixation. For instance, the negative handshape is signed on certain body parts giving rise to items like ^, ^  ^, which mean ‘deaf ’, ‘dumb’, and ‘blind’, respectively, in HKSL (Tang : ). Note that these are non-transparent forms derived through affixation.

.. C

.................................................................................................................................. This overview about the ways of expressing negation in sign languages has widely confirmed that the well-known core properties of negation systems across natural languages are attested in the visual-manual modality as well. However, some modalityparticular ingredients have been identified as well. Among them, the most prominent one is the widespread use of non-manual markers to (co)express sentential negation, with significant but systematic variation in their status across different (types of) sign languages. Another more general aspect that is inherent to the signed modality is the possibility of exploiting simultaneity in the expression of negative markers at the morphological and syntactic levels. However, there is still much ground to be explored, both in the breadth of sign language varieties, but also in the depth of their descriptions. Future research will certainly complete and add to the generalizations that we are able to draw at the current stage of research.

OUP CORRECTED PROOF – FINAL, 2/3/2020, SPi



 

A

.................................................................................................................................. The research in this chapter was partly made possible thanks to the grants awarded to the author by the Spanish Ministry of Economy, Industry, and Competitiveness and FEDER Funds (FFI--P), by the Government of the Generalitat de Catalunya ( SGR ) and by the European Commission (SIGN-HUB H project ). Two anonymous reviewers are gratefully acknowledged for their comments and suggestions.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ........................................................................................................................

NEGATION AT THE SYNTAXSEMANTICS INTERFACE ........................................................................................................................

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ......................................................................................................................

   -       ......................................................................................................................

 . 

.. N-: T  

.................................................................................................................................. T first appearance of Neg-raising in generative grammar took the form of a rule proposed by Charles Fillmore for the “Transposition of NOT (EVER)” within a fragment of generative grammar: “Under certain conditions (e.g. after verbs like want or think which are themselves not negated), a not in the embedded sentence may be moved in front of the main verb . . . ” (Fillmore : ). In motivating both his rule and the syntactic cycle itself, illustrated by the “repeated applications” of this movement rule involved in his proposed derivation of (a) from the structure directly underlying (b), Fillmore did not propose any explicit evidence for a syntactic analysis. He relied instead on the (purported) paraphrase relation between (b) and one reading of the putatively ambiguous (a) (Fillmore : , fn. ). () a. I don’t believe that he wants me to think that he did it. b. I believe that he wants me to think that he didn’t do it. The paraphrase relation assumed here rests on somewhat shaky ground; Dwight Bolinger was widely credited for pointing out (in a  letter to George Lakoff) that the speaker’s negative commitment is weaker in (a) (on its relevant reading) than it is in (b). In this observation, Bolinger was recapitulating the remark of the grammarian Poutsma (: ), who had noted four decades earlier that “the shifting of not often has the effect of toning down the negativing of a sentence.” Putting aside the content of Poutsma’s generalization, note the dynamic metaphor he employs: even thirty years before the introduction of transformational rules, it was natural for a grammarian to conceive of NR in movement terms. After some years of neglect, Fillmore’s syntactic approach to what had come to be called negative transportation or Neg-raising (NR), following the introduction of the NEG element in Klima (), was supported by direct evidence. The most influential argument was conceived by Robin Lakoff (a). Citing an earlier personal communication from Masaru Kajita, Lakoff points out in her CLS  paper that a strict negative polarity item (e.g. until midnight, in weeks) that normally requires a negative licenser within its clause is

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 . 

nevertheless well-formed when embedded under a negated predicate just in case that predicate allows Neg-raising, and furthermore (for some speakers) this structure licenses positive tags formed on the embedded (and by hypothesis negative) clause: ()

a. b. c. d.

I didn’t think John would leave until tomorrow. *I didn’t say John would leave until tomorrow. I don’t suppose the Yankees will win the pennant, will they? *I don’t suppose the Yankees will win the pennant, {won’t they?/do I?}

These contrasts provide support for a grammatical rule of NR within a framework that would account for NPI licensing and tag-formation before the Neg is raised into the main clause. In fact, as we shall see, the functional transparency exhibited by NR predicates—and only NR predicates—that overrides the locality restriction on strict NPIs turns out to be a more complex matter than first appeared.¹ The polarity argument for a syntactic rule of NR is further chronicled in Horn  and discussed insightfully and in great detail by Collins and Postal (); see also van der Wouden (), Israel (b), Giannakidou (), Gajewski (), and Horn () for general considerations on the character and distribution of strict polarity items. (See Chapter  in this volume.) As we shall see, however, the premises on which this argument rests are somewhat shakier than they appeared. If the conception of NR as a syntactic rule dates back to , the generalization for which it accounts has a much longer history. Indeed, unlike most rules of grammar, Negraising has its own patron saint. St. Anselm observes in his argument against Paulus, a Roman jurist six centuries his predecessor, that “non . . . omnis qui facit quod non debet peccat, si proprie consideretur,” that is not everyone who does what he non debet (lit. ‘notshould’) sins, if the matter is considered strictly (i.e. with wide-scope negation, as suggested by the surface structure). The problem, he sees, is the standard use of the string non debere peccare to convey the narrow-scope contrary debere non peccare, rather than the literal wide-scope contradictory (= ‘it is not a duty to sin’). A man who does what is not his duty does not necessarily sin thereby, but (given the interference of the NR reading) to utter (a)—the proposition that a man need not marry—is almost inevitably to commit oneself to the stronger (b), an injunction to celibacy (Henry : ff.; cf. Williams ; Hopkins ; Horn : ). ()

a. non debet ducere uxorem lit., ‘ [he should take a wife]’ b. debet non ducere uxorem lit., ‘he should  [take a wife]’

¹ While Fillmore’s constraint that all predicates intervening between negative licenser and strict NPI must be Neg-raisers may be a necessary condition for grammaticality in sentences like (1a) and (2a), it is not sufficient, as seen in the asymmetry in cyclic application revealed in (i) and (ii), from Horn (: –), an asymmetry for which Gajewski () offers an explanation. i. I don’t believe John wanted Harry to die until Saturday. ii. *I don’t want John to believe Harry died until Saturday.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

-



For Henry (: ), Anselm’s take on modal/negative scope interaction is “complicated by the quirks of Latin usage” and his recognition that “ ‘non debet’, the logical sense of which is ‘It isn’t that he ought’, is normally used not to mean exactly what it says, but rather in the sense more correctly expressed by ‘debet non’ (‘he ought not’).” Henry’s assessment parallels that of logicians and epistemologists seeking to debug English of a similar flaw, from Quine’s dismissal (: –) of the “familiar quirk of English whereby ‘x does not believe that p’ is equated to ‘x believes that not p’ rather than to ‘it is not the case that x believes that p’” as an “idiosyncratic complication” to Hintikka’s complaint (: ) that “the phrase ‘a does not believe that p’ [~Bap in his notation] has a peculiarity . . . in that it is often used as if it were equivalent to ‘a believes that ~p’ [Ba~p]” and Deutscher’s acknowledgment (: ) that I do not believe that p can be “unfortunately ambiguous” between disbelief and simple non-belief. Prescriptive grammarians, too, find NR readings more to be censured than explained; for Vizitelly (), I don’t believe I’ll go and I don’t think it will rain are “solecisms now in almost universal use. Say, rather, ‘I believe I will not go’; ‘I think it will not rain.’” In fact, as demonstrated by half a century of scholarship, Quine’s “quirk”—the lowerclause understanding of higher-clause negation over a semantically coherent but somewhat variable range of predicates and operators—is far more than an unfortunate foible or error in English (or Latin) usage. But just how is it to be accounted for?

.. M- 

.................................................................................................................................. “The essence of negation,” proclaimed the nineteenth-century neo-Hegelian philosopher Bernard Bosanquet (: ), “is to invest the contrary with the character of the contradictory.” One prime instance of this investment practice is litotes (the fact that “from ‘he is not good’ we may be able to infer something more than that ‘it is not true that he is good’ ”—Bosanquet : ) “ . . . the habitual use of phrases such as I do not believe it, which refer grammatically to a fact of my intellectual state but actually serve as negations of something ascribed to reality . . . Compare our common phrase ‘I don’t think that’— which is really equivalent to ‘I think that ___ not’ ” (Bosanquet : ). Over the years, empirical and theoretical considerations have gradually led linguists away from Fillmore and Lakoff (and Horn ) and down the trail blazed by Jackendoff (: ), who points out that Neg-raising interpretations are more accessible with firstperson than with third-person subjects, which would be an odd property for an actual rule of grammar to manifest. Indeed, he maintains, “The synonymy between John thinks that Bill didn’t go and one reading of John doesn’t think that Bill went is inferential in character and has nothing to do with the syntactic component—it may even have nothing to do with the semantic component.” (This analytic trend is reviewed in Horn  and Horn : ch. .) The result has been a de facto rejection of the syntactic approach to NR. Two decades after Fillmore proposed his rule, Lerner and Sternefeld (: ) characterized Neg-raising as “Eine Transformation aus dem goldenen Zeitalter”—a transformation from the Golden Age; more recently, Levinson (: ) took it as noncontroversial that “the syntactic analysis [of NR] no longer fits with today’s syntactic theories.”

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 . 

While, as we shall see, this conclusion has since been vehemently challenged, it is worth revisiting the positive grounds on which the NR phenomenon has been seen as an instance of a pragmatic pattern. Standard grammars of English tend to subsume NR (if it’s noticed at all) under the general tendency for formal contradictory negation paraphrasable as ‘It is not that case that p’ to be strengthened to a contrary (the “MaxContrary” conspiracy of Horn b), as when She’s unhappy or even She’s not happy are understood as making a stronger negative claim than a mere denial that she is happy. These lexical or syntactic negatives can be viewed as expressing lexicalized and virtual (or on-line) contrariety respectively. And as Bosanquet notes, the effect of contrary or strong readings for negated clause-embedding believe is paralleled by similar strengthened readings for monoclausal believe, where NR movement is excluded. ()

a. I don’t believe in {you/your story}. b. I don’t believe that your story is true.

()

a. I don’t believe {in God/there is a God/that God exists}. ( view) b. I believe {there is no God/that God does not exist}. ( view) c. I don’t just not believe in God. I believe there is no God. (, April , )

Following Horn (: ch. ), Huddleston and Pullum (: ch. , §) note that negative clauses often “are taken to be making a stronger claim than they actually entail”; NR is cited (without being named) merely as “a further special case of this tendency to move to a more specific interpretation of a negative clause: the case where a clause containing a clausal complement has negation in the matrix clause interpreted as applying to the subordinate clause.” In this respect, Huddleston and Pullum partially echo Jespersen (: ), who presents the use of I don’t think to “really mean” I think he has not come as both an instance of specialization of negative meaning and an illustration of “the strong tendency in many languages to attract to the main verb a negative which should logically belong to the dependent nexus.” But unlike Jespersen (or Poutsma  on the “shifting of not” as noted above), Huddleston and Pullum make no allusion to “attraction” or movement of the negative element. For Jespersen, the tendency in question is part of a more general conspiracy in natural language to signal negation as early as possible. This Neg-first principle (Horn : , after Jespersen : ) has been held responsible for diachronic shifts in the expression of sentential negation (the “Jespersen cycle” of Dahl ; cf. van der Auwera  for a recent overview) and the emergence of ambiguities arising in contexts like [ S  S] (Jespersen : ), as in “He didn’t marry her(,) because he really cares for her,” where the “illogical” scope reading—on which his affection was the non-cause of the wedding rather than the cause of the non-wedding—can be rendered more or less accessible by the intonation contour. However, Neg-first is a somewhat dubious motive for NR, given the possibility of lower clause understandings for higher clause across verbs glossed as ‘think’, ‘believe’, or ‘want’ in verb-final languages, where the effect of NR is movement up and to the right rather than up and to the left. Such readings are available in Korean (as in (), from Lee and Hong ) and Turkish (as in (), from Kennely , Süleyman ), although in the latter case such readings are limited to non-finite complements:

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

-



()

a. Chelswu-nun [Mina-ka an o-ntako] sayngkakha-nta. Chelswu- Mina-  come- think- ‘Chelswu thinks that Mina is not coming.’ b. Chelswu-nun [Mina-ka o-ntako] sayngkakha-ci anh-nunta. Chelswu- Mina- come- think- - ‘Chelswu does not think that Mina is coming.’

()

a. Ben [bu oda-da kimse-nin sigara iç-me-di-k-in]-i I this room- anybody- cigarette smoke----- zannet-iyor-um. think-- ‘I believe that nobody smoked in this room.’ b. Ben [bu oda-da kimse-nin sigara iç- di-k-in]-i I this room- anybody- cigarette smoke- --- zannet-m-iyor-um. think- -- ‘I don’t believe that anybody smoked in this room.’

But while NR in such verb-final languages fails to correlate with Neg-first, it does exhibit the attenuation or politeness motivation associated cross-linguistically with NR. How does the weaker contradictory, I don’t believe that p understood as ¬(I believe that p), manage to strengthen to the contrary I believe that ¬p? The classic recipe is given by Bartsch () in a manifesto for the non-existence of NR that trumpets its conclusion in its title: “ ‘Negative transportation’ gibt es nicht.” Like Jackendoff and other NR-skeptics, Bartsch (: ) rejects any ambiguity for (), but argues that mirroring the semantic entailment from (a) to (b), there arises in certain contexts a pragmatic implication from (b) to (a), arising from the assumption that the subject a can be assumed to have given some thought to the truth of p and come to some conclusion about it, rather than that a hasn’t thought about p or is neutral as whether p or ¬p. ()

Peter glaubt nicht, daß Hans kommt. ‘Peter doesn’t believe that Hans is coming.’ a. Peter glaubt, daß Hans nicht kommt. ‘Peter believes that Hans is not coming.’ b. Es ist nicht so, daß Peter glaubt, daß Hans kommt. ‘It is not the case that Peter believes that Hans is coming.’

Bartsch takes propositional attitudes (‘think’, ‘believe’, ‘want’) to express the subject’s cognitive or psychological stance toward the complement, inducing a disjunctive pragmatische Voraussetzung (pragmatic presupposition) of the form “[a believes that p] or [a believes that ¬p].” Thus “Neg-raising” is not a rule of grammar or semantic interpretation but a (mere) pragmatische Implikation; while (a,b) are semantically distinct, they will tend to express the same information relative to a given Sprechsituation. Bartsch’s inference schema can be represented as in (): ()

(i) F (a, p) v F (a, ¬p) the pragmatically presupposed disjunction (ii) ¬F (a, p) the proposition actually asserted (iii) F (a, ¬p) the proposition conveyed

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 . 

The key is the assumed disjunction in (i): if you can assume I either want to go or want to stay (= not-go) and I say (out of diffidence, politeness, cowardice, etc.) ‘I don’t want to go’, you can infer that I want to stay: From the assumed disjunction p or q and the expression of not-p, infer q. This is the logical schema variously known as disjunctive syllogism, modus tollendo ponens, reasoning by exclusion, or (if you’re a Stoic) the fifth indemonstrable syllogism. More recently, Klooster (: ) adduces a “black and white” effect to account for the strengthening of contradiction to contrariety with Neg-raising predicates: In a discourse where judgements and intentions are relevant, but reserving or deferring them are not, verbs of the considered type are easily interpreted as dichotomous . . . In a sentence containing a matrix verb of the type in question, its contrary can thus (indirectly) be expressed simply by introducing negation. That is, where P is an NR verb, x the subject, and p the complement clause, the following seems to hold: ¬P(x, p) iff P(x, ¬p).

The problem, as Klooster recognizes, is that this proposed solution to NR cannot handle variation within and across languages as to just which NR candidates can substitute for “P” in this formula—or for “F” in the derivation in (). When is the disjunction in (i) actually assumed? Not just any predicate of the knowledge and belief class will do. In fact, membership in the class of propositional attitudes is neither necessary (given Neg-raisers like should, ought to, be likely, or be advisable) nor sufficient (given non-Neg-raisers like know, realize, or be sure) for “dichotomous” behavior. Just which predicates license this lower-clause understanding of higher-clause negation? One property common to all NR triggers not touched on by Bartsch is the relative slenderness of the functional difference between the pre-raised form with lower negation and the logical form with the upstairs negative taking wide scope. be able be possible ←WEAKER— may, might can, could allow, permit, let be allowed be legal

—STRONGER→

believe, suppose, think be likely, probable figure to seem, appear, look like

know, realize be clear, evident be sure, certain be odd, significant

should, ought to, better be supposed to be desirable, advisable be a good idea want, choose, intend, plan to suggest, advise

must, have to need, be necessary be obligatory make, cause, force order, demand, require

It is the closeness of the external (contradictory) readings of not likely, not believe, and not advisable to likely not, believe not, and advisable not, respectively, which allows these predicates to function as Neg-raisers, and the more significant distance of not possible, not realize, and not obligatory from possible not, realize not, and obligatory not which removes these predicates from that category. Neg-raising licensers tend to be positive mid-scalar operators, situated near the midpoint of the relevant positive scale (Horn , ). Predicates and operators that are too strong (know, be certain, always, emotive factives like regret or surprising), too weak (be able, be possible, allow, can), or inherently negative (doubt, deny, prevent, never) are never Neg-raisers. This can be seen exhibited in the contrasts in () between those elements that do and those that do not allow NR readings in English (see Horn  for elaboration):

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

-



() believe, suppose but not know, doubt, disbelieve want, suggest but not insist, forbid, prohibit advisable, desirable but not obligatory, forbidden should, ought to, better (I don’t think you should leave) but not have to, may, can likely, probable but not certain, impossible most (You don’t know most of the people here) but not all, many, few usually (It doesn’t usually snow in May) but not always, often, sometimes, rarely As seen in the examples in (), the mid-scalar generalization cross-cuts the traditional distinction between lexical and functional operators, as exemplified by the appearance of stative verbs, modal auxiliaries, adjectives, determiners, and quantificational adverbs among the Neg-raisers, provided they fit the scalar requirement.² Another striking property of the distribution of these items is their polar asymmetry. As has often been noted (see Prince , or recall Poutsma’s “toning down the negativing of a sentence”), Neg-raising can be seen as a euphemism attenuating negative force, predicting the absence of inherently negative Neg-raisers. The strengthening of contradictory negation to contrariety induced by NR is equally on display in litotes. As Ducrot () has observed, a negation of a positive predication as in (a) is typically intended and understood as a contrary, while a negation of the corresponding negative assessment as in (b) expresses a simple contradictory. ()

a. Pierre n’est pas gentil. ≈ Pierre est méchant.

‘Pierre isn’t nice.’ ‘Pierre is naughty.’

b. Pierre n’est pas méchant. ‘Pierre isn’t naughty.’ ≠ Pierre est gentil. ‘Pierre is nice.’ The negation of a favorable (unmarked) term typically conveys the affirmation of an unfavorable (marked) term via R-based implicature, but not necessarily vice versa (see Horn , a for elaboration). This asymmetry is attributable to politeness (“Respect negative face”; Goffman ; Brown and Levinson ), suggesting a maxim applying to both litotes and NR: “If you have something negative to say, don’t say it directly.”³

² In many languages the midscalar constraint can be overridden in the case of strong deontics whose negation—as in the case of Tobler’s ne pas falloir—allow lower-clause understandings; see Horn :  for discussion. ³ The weakening effect of NR can be seen as an instance of the general tendency for negative force to be inversely correlated with the distance of the negative marker from its target (see Horn : ). Chris Collins (p.c.) points out that in many cases no true politeness is signaled by NR, given the naturalness of sentences like (i). (i) I don’t fucking think he has seen his mother in years. Such examples may seem to disprove the claims by Poutsma, Bolinger, Prince, et al. associating NR with politeness, euphemism, or “softening,” but can arguably be attributed to conventionalization. A similar case is that of indirect speech acts, usually taken to be more polite or euphemistic than the corresponding direct speech act (compare ‘Close the window’ to ‘Could you close the window?’ or to ‘Would you be so kind as to close the window?’). Here too the existence of examples like ‘Would you be so kind as to close

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 . 

But on its own, the pragmatic factors we have discussed cannot suffice for NR, as pointed out by the great Romance philologist Adolf Tobler () in his eponymous paper on Il ne faut pas que tu meures. The preferred meaning—and for some speakers the only meaning— of this sentence in modern French is ‘You mustn’t die’, amounting to a “logisch ungerechtfertigte Stellung” (‘logically unwarranted position’) of negation, since the higher-clause negation would be expected to yield the wide-scope negation ‘It is not necessary that you die’. Tobler cites a variety of German and French verbs reflecting this same pattern, including German wollen ‘want’ and sollen ‘should, ought to’ and their (older and modern) French counterparts vouloir and devoir, and contrasts the resultant (NR) readings as in () with the straightforward instances of ironic negation in (). ()

a. Il ne faut pas qu’elle entre. it  be-necessary  that-she enter- ‘She must not enter.’ b. Je ne veux jamais que son nom soit prononcé devant moi. I  want (n)ever that his/her name be- pronounced before me ‘I want her name never to be pronounced in front of me.’ [lit. ‘I don’t ever want her name to be pronounced in front of me’] c. Un peintre ne doit penser que le pinceau à la main. a painter  should think  the brush at the hand [ne . . . que = ‘only’] ‘A painter should think only with brush in hand.’

()

the holy saints by whom God was not hated (= was loved) he will do you little good (= all the harm he can) we weren’t in the best possible situation (= in the worst situation) he didn’t show that he had forgotten their kindness (= he showed he hadn’t forgotten)

As Tobler points out, the conscious irony or litotes accompanying the interpretation of the examples in () is absent from the interpretation of their more conventionalized counterparts in ().⁴ And then there is the problem of variation within and across languages. Necessary conditions for Neg-raising are not sufficient ones. While hoffen Neg-raises, Eng. hope

the fucking window’ indicates that this sort of politeness can be overridden, although such a case wears speaker pretense or irony more obviously on its sleeve than the corresponding NR case does. ⁴ Tobler’s own analysis of the true NR examples in (9), drawing on the sense that they involve a kind of fusion of the two clauses with the negation affecting the semantics in both, anticipates later developments: [T]he idea that in I don’t think he’s coming we have a negative element that belongs truly to the subordinate verb and that can be transferred, like a syntactic ping-pong ball, to another position without altering its logical connections, I think is not quite true . . . It does not merely hop from one [clause] to another but belongs semantically to both. (Bolinger : –) [W]e are used to thinking of a negative as occurring only on one specific S for each act of negation; but perhaps in the ‘negative transportation’ sentences . . . there is a requirement that each clause be under the umbrella of a negation. (Cattell : )

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

-



(usually) does not; Latin sperare ‘hope’ Neg-raised but French espérer doesn’t (while souhaiter ‘wish, hope’ does). Parenthetical guess is a Neg-raising propositional attitude in Southern U.S. English but not in other U.S. or U.K. varieties. And the predicate most clearly transparent to negation is not in fact a Neg-raiser, as seen from the contrast in () (Horn : ; Collins and Postal : ). ()

a. It is true that John will not {leave until tomorrow/do a damn thing}. b. *It is not true that John will {leave until tomorrow/do a damn thing}.

The clear conclusion is that no purely semantico-pragmatic account that does not acknowledge some form of conventionalization can successfully deal with the NR phenomenon. Different responses to this conclusion have been proposed. Horn and Bayer  (see also Horn : ch. ) invoke the notion of conventions of usage (short-circuited conversational implicatures). For Gajewski (), the Bartschean excluded middle is an Abusch-style soft presupposition (cf. Abusch ). Romoli () takes NR to be a scalar implicature. While there may be useful semantic generalizations about which predicates potentially license NR, not all members of the suitable classes are Neg-raisers in a given language. As Lauri Carlson (: –) puts it, there is on the one hand a “general tendency of any expression of doubt or indecision to suggest disbelief or disinclination” but we must acknowledge on the other hand that “the lexical selectivity of the negative transportation phenomenon does suggest that whatever process creates it has become a conventional rule in the clear cases.” Thus, the negations in I don’t want that toy and I don’t want to go may both represent contraries in contradictory clothing, as do the lexicalized contraries resulting from prefixal negation (uncommon, unhappy, infamous, impious), but these construction types differ from each other in the degree to which pragmatic strengthening has become conventionalized, a matter that casts doubt on purely meaning-based approaches without serving as decisive evidence for a syntactic theory of NR. Bartsch’s schema for pragmatic strengthening through disjunctive syllogism, while empirically flawed as a model of NR (a problem addressed in recent neo-Bartschian accounts by Gajewski  and Romoli ), in fact functions as a useful template for other cases in which formal contradictories are massaged into acting contraries via assumed disjunctions; see Chapter  in this volume. But if a purely meaning-based account is unworkable, what are the prospects for a grammatical alternative? Might the reports of its death turn out to be grossly exaggerated?

.. N-  : T  

.................................................................................................................................. Despite the evidence we have reviewed suggesting that no purely meaning-based treatment of the Neg-raising phenomenon is empirically adequate, it has long appeared that the Despite the vividness of ping-pong balls and umbrellas as metaphors for negation, such fusion analyses have proved remarkably resistant to formalization, although see now Collins and Postal () for a syntactic analysis in a similar spirit.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 . 

grammatical approach to NR is even more flawed. Among the other difficulties ensnaring such an approach, a seemingly decisive problem is posed by data like that in () (cf. Horn : –; : ). ()

Nobody supposes that peace in the Middle East is possible. Neither the Democrats nor the Republicans think that he’s telling the truth. {Neither/None} of them would be likely to date you. I drank more than I should have. Only Chris ought to know about it. You’re never supposed to get that drunk at a funeral. {Few/Scarcely any} of my friends think you’d lift a finger to help. I doubt that he wants to attend your talk.

In each of these cases the italicized element is semantically associated with the clause below the (boldface) NR predicate—suppose, think, likely, should, etc.—while being lexically incorporated into a quantifier, connective, adverb, or verb in the higher clause. Thus, the first example is taken to indicate that everybody supposes Middle East peace is not possible, the second that both Democrats and Republicans think he’s not telling the truth, and so on. For decades, this consequence of adopting a grammatical program for NR amounted to a reductio for those who subscribe to the widely received position that rules of word-formation (as opposed to rules of inflectional morphology, on some accounts) cannot apply to the output of syntactic rules.⁵ As recently as Homer (: –), the consensus view has seen negative incorporation as a deathknell for a grammatical rule of NR, especially combined with the earlier positive arguments for meaning-based approaches, whatever the differences among these approaches and whatever the acknowledged problems stemming from lexical idiosyncrasy (cf. Horn ; Gajewski ; Romoli ; Homer ). Rowing against this semanticopragmatic current, and on the golden anniversary of Fillmore , Chris Collins and Paul Postal have mounted an impressively rich and detailed defense of “Classical Neg-raising” as a rule of grammar. In a monograph and a series of papers (Collins and Postal , , a, et al.), they have made the case that only a syntactic account can do justice to a variety of empirical arguments that pose embarrassing if not insurmountable difficulties for meaning-based accounts of NR while being compatible with their syntactic program. These include the apparent island-bound character of NR (: chs. –), the distribution of pleonastic negative parentheticals (: ch. ), and especially the restriction on longdistance strict polarity items under non-clausemate negation to contexts in which intervening predicates are all NR licensers (essentially the Lakoff a argument reviewed in section . above adumbrated with additional empirical and theoretical support). At the same time, they respond to the challenge of what they call the “composed quantifier argument” from Jackendoff (: ), Horn (: ; : –), Gajewski (: )

⁵ Indeed, the existence of the pattern in (15), if not its theoretical implications, was recognized by St. Anselm, who observes that a sentence of the form Solus homo debet facere X (‘Only (a) man ought to do X’) tends to be read as asserting that Quisquis non est homo debet non facere X (‘Any non-man ought not to do X’): “The negative force implicit in the applicative ‘only’ can extend, so to speak, beyond the main verb of the clause in which it appears to infinitives governed by the main verb” (Williams : ).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

-



et al. based on examples like those in (), without accepting the intolerable assumption of post-syntactic lexical rules. Instead, they posit a more abstract syntactic analysis featuring a second (unraised) negative operator in the matrix clause and a deletion rule eliminating one Neg operator subsequent to NR (Collins and Postal : ch. ). Following Horn (, ), Collins and Postal (: –) go on to construct an ingenious argument for the syntactic nature of NR based on the availability of embedded subject-auxiliary inversion under a fronted NPI licensed by a higher negation, as in (a), uttered by a reporter in reference to the  abduction of Patty Hearst by the Symbionese Liberation Army. ()

a. I don’t think that ever before have the media played such a major role in a kidnapping. b. I think that  [the media have (at) some time before played such a major role in a kidnapping]

Collins and Postal endorse the observation of Horn (: ; : –) that while the inversion in the well-formed (a) can be taken to derive from a structure like that in (b), no such inversion is available in the absence of NR predicates, whence the impossibility of replacing think in (a) with claim or regret. McCawley (b: ) provides the additional (constructed) examples in (), which Collins and Postal supplement with a broad array of attested and constructed examples including those in () to illustrate what they term “Horn clauses” (henceforth H-clauses): ()

a. I don’t suppose that under any circumstances would he help me. b. We didn’t anticipate that at any time would our work create difficulties.

()

a. I don’t believe that at any time did traffic come to standstill. b. I didn’t expect that for any reason would she agree to that. c. I didn’t imagine that either of them would she be anxious to marry.

For Collins and Postal (), as for Horn () and McCawley (b), the acceptability of embedded H-clause inversion under higher (non-tautoclausal) negation, like that of strict NPIs (recall (a) above), requires the presence of a Neg-raising trigger and thus motivates a grammatical analysis of the phenomenon. The problem with this otherwise elegant argument for a syntactic rule of NR is that embedded subject-aux inversion following a fronted NPI is in fact possible with certain non-Neg-raising predicates, as revealed by the possibility of H-clauses under negated non-factive know, a staunch non-Neg-raiser. Attested sentences like those in () are provided in Horn (): ()

a. I don’t know that ever before or since in my life have I felt such exhaustion. (Hugh Walpole, The Dark Forest, ) b. I don’t know that EVER before had all three boys napped simultaneously. () c. I don’t know that ever again will we have four – guys on the team at the same time. ()

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 . 

H-clause inversion is likewise possible under can’t say and (non-factive) not aware, which equally block classical Neg-raising, as seen from the non-equivalences in (): ()

I don’t know that p ≠ I know that not-p I can’t say that p ≠ I can say that not-p I’m not aware that p ≠ I’m aware that not-p

As above, the googled examples in () are fully acceptable in my own dialect—and in Collins and Postal’s: ()

a. I can’t say that at any time did I have a problem with any of the customer service team. () b. I can’t say that at any time did I feel these headphones were doing any sort of active noise cancelling of the airplane’s cabin. () c. I’m not aware that ever in the history of New York State has something like this happened. It’s an unfortunate first. ()

Note that although no lower-neg paraphrase is available in such cases, there is a negative proposition that is at issue in such cases; in (a), for example, the speaker’s point is to indicate that he suspects that he has never had a problem with customer service team members. Crucially, this same class of non-Neg-raisers tends to allow strict NPIs in embedded clauses, contra the received wisdom, and again with the implicit suggestion that the negation of the embedded proposition holds, so that Isabel, the author of (c), is understood as suggesting that she hasn’t cooked herself a full meal in weeks. ()

a. I don’t know that Santa comes around these parts until Christmas Eve. () b. [W]ith my 1st pregnancy . . . I don’t think I got breast milk. I don’t know that I have it this time either. () c. I can’t say I’ve cooked myself a full meal in weeks, if not months. () d. I am not aware that workplaces are all that different in Montana than elsewhere in the US. ()

(The fact that non-clausemate negation can license embedded until phrases under strong implicature in the absence of NR was earlier noted by Lindholm  and Baker .) To account for the broader range of contexts allowing H-clause inversion and strict NPIs as observed in Horn (), Collins and Postal (a) have adjusted their syntactic theory of Neg-raising to deal with what they term “CU predicates” (for “Cloud of Unknowing”, à la Horn ). While a “standard” H-clause with a “dichotomous” predicate (to adopt Klooster’s term) as in (a) [= (a)] can be represented as in (0 a), the corresponding version with know in (b) must, to yield the correct semantics, derive from a structure like

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

-



(0 b), in which “there are three semantically and syntactically distinct NEGs,” two of which are covert and eventually deleted before the surface string emerges (Collins and Postal a: –). ()

a. I don’t think that ever before have the media played such a major role. b. I don’t know that ever before have the media played such a major role.

(0 )

a. I 1 think that [[  ever before] have the media . . . ] b. I [1 know- that [[ [[ ever] before] have the media . . . ]]]

Similarly, ()—a line from the  Clint Eastwood western, Unforgiven—derives from (0 ): ()

Hell, I don’t know that it was all that easy even back then.

(0 ) I [ know- that [[ [it was [ all that easy] even back then]]] But while it’s always possible to enrich a syntactic account of Classical (or neo-Classical?) Neg-raising, and while the account provided in Collins and Postal (a) is as ingenious as it is baroque, the danger is that proliferating silent Neg operators is ad hoc. Furthermore, it will be noticed that the structures Collins and Postal (a) ultimately endorse for dichotomous predicates (I don’t think that . . . ) and “CU” predicates (I don’t know that . . . ) are fundamentally distinct. It has also been argued that the adoption of a syntactic account of NR may have undesirable consequences elsewhere, for example forcing the view that VP ellipsis must require formal and not semantic identity (Jacobson ). Syntactic accounts, whatever their details, run the risk of overlooking or minimizing the properties NR shares with (other) instances in which apparent contradictories are strengthened (by the hearer) to contraries, either lexically (She’s unhappy) or pragmatically (He’s not happy). In each case, as we have noted (modulo the complications noted in fn. ), the typical motivation for the speaker to weaken the negative force is politeness (“Respect negative face”), whence the vast preponderance of first person, present tense citations for the CU cases (and to a large extent, as first noted in Jackendoff , for the NR ones as well). These are all means to express a belief that ¬p while appearing to refrain from doing so. Negated CU predicates constitute an even more polite or euphemistic way to express ¬p with weakened force than do negated NR predicates, given scale reversal: ()

P : < I believe that p, I know that p> N : < I don’t know that p, I don’t believe that p>

Note that neither ignorance nor committed agnosticism toward ¬p is compatible with strict NPI licensing with either dichotomous or “CU” verbs:⁶ ⁶ The incompatibility of agnostic sentences like (26) with negation is also discussed by Collins and Postal (2004: §16.6). Note that to not know whether signals full agnosticism: If I don’t know whether p, then I don’t know that p and I don’t know that ¬p. So the whether counterparts of sentences like (19) or (22) are predictably ruled out:

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 . 

()

a. I don’t think he’s getting here until 1:00 (#and I don’t think he isn’t—who knows?). b. I don’t believe that ever before have the media played such a major role (#and I don’t believe they haven’t; I haven’t really thought about it one way or the other). c. I don’t know that they’ve lived here in years (#and I don’t know that they haven’t). d. I don’t know that Santa comes around until Christmas Eve (#but I think he does).

One more challenge for both the syntactic and the neo-Bartschean excluded middle account of Gajewski (): We have seen that the majority of dichotomous (non-CU) NR predicates are midscalar statives, with a smattering of strong deontics like French falloir. What both the syntactic account and the neo-Bartschian fail to predict is that Negraising predicates can never be weak scalar values. Thus, in no language do predicates like allow, be able, be possible, be permitted, can, et al. trigger NR readings, despite the fact that the excluded middle holds in such cases, as in (): ()

a. [Lee was able to leave] ∨ [Lee was able not to leave] b. [It is possible that it will rain today] ∨ [It is possible that it will not rain today]

Crucially, the higher-neg reading (e.g. Lee wasn’t able to leave) would be stronger, not weaker, than the lower-neg reading (Lee was able not to leave). In effect, the disjuncts in the Bartschean formula must be mutually exclusive, assuring that NR involves derived contraries (which can be false but not true together), not derived subcontraries (which can be true but not false together). In the terms of our earlier discussion, while NR does not necessarily participate in the Neg-first conspiracy (cf. (), () above), it is complicit in the MaxContrary conspiracy, the tendency to maximize contrary readings of apparent contradictory negation. Given our focus in this section on the potential problems for a syntactic account of NR, it seems only fair to conclude with a case in which such an account seems to be on sounder empirical footing than the corresponding semantic or pragmatic approaches. We have seen that the success of the arguments for grammaticalizing NR based on the distribution of strict NPIs and on the availability of H-clause inversion are weakened by the fact that the same phenomena are available for CU predicates (know, be aware, can say) whose negations do not license ordinary NR-style dichotomous interpretations. But now consider the distribution of pleonastic negative parentheticals (PNPs). Affirmative utterances permit positive and doubly-negative parentheticals, but not simple negative ones, in either sentence-final or medial position, as seen in () and (): ()

a. The country, I {heard/believe/think/know/understand}, will recover. b. *The country, I {don’t believe/don’t think/doubt/deny}, will recover. c. The country, I {don’t doubt/don’t deny}, will recover.

(i) I don’t know {that/#whether} Santa comes around these parts until Christmas Eve. (ii) I don’t know {that/#whether} ever before have the media played such a major role.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

- ()



a. The country, everybody {believes/knows}, will recover. b. *The country, nobody {believes/knows}, will recover. c. The country, nobody {doubts/denies}, will recover.

But as shown in (), negation is sometimes permitted pleonastically within parentheticals (the two versions of (b) are synonymous) after the negative force is marked by main clause negation. ()

a. Robin, I (*don’t) think, left on time yesterday. b. Robin didn’t, I (don’t) think, leave on time yesterday. c. Robin, I (*don’t) think, didn’t leave on time yesterday.

Crucially, as first pointed out by Ross (), the PNP in (b) is only possible with certain predicates, as seen in (): () a. Robin won’t, I don’t {think/believe/suppose/anticipate/expect}, get here on time. b. Robin won’t, I (*don’t) {claim/fear/maintain/insist/regret}, get here on time. Based on contrasts like this, Ross observes: “What is significant about the class of verbs which can appear in negative parentheticals is that it is precisely the class of verbs which undergo Not- [= NR]” (Ross : ).⁷ Ross’s generalization doesn’t quite work, however, for reasons anticipated by Bolinger (). For one thing, NR-class membership is not sufficient for generating a grammatical PNP (or indeed any parenthetical): *Robin won’t get here on time, I don’t want. For another, “negative parentheticals” in Ross’s generalization must be limited to PNPs, given the possibility of doubly negated parentheticals like those in (c) and (c). Nor is the formulation of just which structures are allowed in PNPs entirely straightforward (see Collins and Postal : ch.  and Collins and Postal a for elaboration). But the key point in Ross’s generalization is correct: with the appropriate provisos, the possibility of PNPs really is limited to true dichotomous NR predicates. Crucially, the same predicates that hurl monkey wrenches into the mechanisms of strict polarity and H-clause arguments are ruled out in PNPs: ()

a. Your analysis of NR isn’t convincing, {I don’t think/I don’t believe/it doesn’t seem}. b. *Your analysis of NR isn’t convincing, {I don’t know/I can’t say/I’m not aware}.

More generally, “CU-sentences never permit corresponding negative parenthetical clauses” (Collins and Postal a: ).

⁷ Ross captures this generalization by breaking up NR into a two-stage process of copying and deletion. In the derivation of sentences like (31a), parenthetical formation (or ‘S[entence]-lifting’) would bleed the deletion operation, while if Slifting fails to apply, the outcome is identical to that of standard NR, e.g. I don’t think Robin will get here on time. See Ross (: –) and Collins and Postal (: –) for elaboration.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 . 

We seem to have arrived at an awkward impasse with signposts pointing us in two incompatible directions, each of them favored by certain considerations and disfavored by others, as summarized by a roster of advocates for each option (Table .).

Table . Advocates for each of the two approaches Syntactic approaches

Semantic/Pragmatic approaches

Fillmore  R. Lakoff a Lindholm  G. Lakoff b Horn  Ross  Seuren  Horn  Horn  (in places) Collins and Postal  Collins and Postal  Collins and Postal a

Jackendoff  Bartsch  Cornulier  Horn  (in places) Horn and Bayer ; Horn : ch.  Gajewski  Romoli  Horn a, a, b Homer  Zeijlstra a Hoeksema  Jacobson 

But what if we not only don’t need to choose between these unitary positions, but must somehow opt for both, even within a single language? Collins and Postal (b) have recently argued that this is indeed the case. After revisiting the previously traveled terrain (NPI licensing, H-clauses, islands, parentheticals), they conclude their study with these remarks: This paper has expanded on previous work, including our own, to argue that neither an exclusionist syntactic approach (based on NEG raising) nor an exclusionist pragmatic/ semantic approach based on EM [Excluded Middle] inferences can provide viable accounts of the known domain of NEG scope fixing. We have thus in effect argued that there are two distinct sorts of phenomena within the NEG scope fixing domain, although the two subdomains overlap extensively, e.g. in the complements of CNRPs [= Classical Neg-raising predicates] which are not islands . . . [E]ven if one subscribes to the CP () arguments that a syntactic account of NEG scope fixing is required for some cases, one should recognize that a unitary exclusionist account of the facts is not viable. (Collins and Postal b: )

For Collins and Postal (b), then, the facts of English demand a two-headed theory that requires invoking a syntactic rule of NR for some cases, a meaning-based conception of NR in others, and overlapping derivations in still others.⁸ To some NR-history buffs, this measured conclusion will ring a few bells, however faintly. Bošković and Gajewski (: §) distinguish DP languages from NP languages as determined by a convergence of morpho-syntactic criteria; they maintain that only the former kind of language allows grammaticalized Neg-raising; NP languages lacking articles (e.g. Serbian/Croatian, Russian, Japanese, Chinese) allow only “pragmatic” Neg-raising when the excluded middle condition is satisfied. Such languages thus exclude the ⁸ Some readers may be reminded of parallel architecture theories like those of Sadock () or Jackendoff ().

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

-



long-distance licensing of strict NPIs under negated NR predicates, a diagnostic of grammatical NR. Even farther back, forty years before the preparation of this chapter, a -page monograph on the topic of Neg-raising concludes with the question “Is NR a syntactic, semantic; or pragmatic phenomenon?” and the response “, guilty as charged, on all counts, pending appeal to a higher court or the washing ashore of decisive new evidence” (Horn : ). Although this answer may seem unsatisfying and tentative in the extreme, and despite the accretion of a variety of apparently decisive new evidence on all sides of the question, it will have to do.

A

.................................................................................................................................. I am grateful to Jack Hoeksema, Polly Jacobson, Pierre Larrivée, Hedde Zeijlstra, and especially Chris Collins and Paul Postal for comments and challenges that have sharpened my thinking on these issues. Earlier versions of some of this material were presented at conferences in Barcelona, Amsterdam, Bochum, and College Park, MD; I am grateful to the audiences at those presentations for feedback.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ......................................................................................................................

    ......................................................................................................................

 

.. T    

.................................................................................................................................. N languages allow for dependencies between linguistic expressions over a distance (see Chomsky  et seq.). In the declarative sentence in (a) all linguistic material occurs in its canonical position. In the wh-interrogative in (b), however, the wh-adverbial how is fronted to the beginning of the clause. Now, the intuitive paraphrase of (b) is something along the lines of Which manner x is such that Mary fixed the car in x? That is, the meaning of (b) has a bound variable x in what corresponds to the canonical position of how, indicated by underline, bound by the wh-expression. Without a variable in the canonical position, it is not clear at all how to capture our intuitions regarding the interpretation of (b). ()

a. Mary fixed the car quickly. b. How did Mary fix the car _? c. What did Mary fix _ quickly?

It is thus tempting to make the overt fronting of the wh-word crucial in establishing the dependency observed in the meaning. Indeed, this is the standard approach to wh-movement and its interpretation going back at least to Karttunen (). Parallel considerations apply to (c), where the wh-argument what undergoes wh-movement establishing a long distance relationship with its canonical position. Such dependencies are often subject to locality restrictions (e.g. Ross ). The effects of locality restrictions can be described in the following way. Two expressions α and β in sentence S need to enter into a relation Ri. The presence of an element ω c-commanded by α and itself c-commanding β, however, blocks the instantiation of Ri, whereby S is perceived to be degraded.¹ Schematically, in (a) α and β, on the one hand, undergo a ¹ At this point the question arises whether the syntactic notion of c-command is the one relevant for the definition of intervention or whether it is the semantic notion of scope that is crucial. These two notions often coincide and it is thus not always easy to discriminate (cf. Barker  a.o.).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



successful relation, depicted by co-indexation and the checkmarks. A similar relation is, on the other hand, prohibited in (b)—depicted by the crosses—by the presence of the intervening element ω. The result of this prohibited dependency is that the sentence is degraded. () a. [ . . . α✓i . . . [β✓i . . . ]] b. [ . . . α✗i . . . [ω . . . [β✗i . . . ]]]

(dependency intact) (dependency broken)

To clarify, α in (b,c) would be the dislocated wh-expression and β its canonical position. Here I tacitly assume that the canonical position is filled by a trace or copy of the fronted material (see Fox , Sauerland  for pertinent discussion). In other words, there is linguistic material in the positions marked by an underline. There is, however, no intervener ω in these examples. One well-known class of restrictions on dependencies is induced by so-called strong islands (Ross ). There is consensus in the literature that strong islands are to be explained by syntactic mechanisms. The degradedness caused by certain other restrictions on dependencies are referred to as intervention effects. On the one hand, intervention effects share with strong islands the general characteristics of restrictions on dependencies outlined above. On the other hand, they differ from them in that not all expressions of a class of potentially harmful interveners is harmful for all kinds of dependencies of a given class. That is, intervention effects are characterized by a certain degree of selectivity: ()

Intervention effect Given classes of linguistic expressions A, B, and X, an intervention effect occurs in sentence S if some but not all configurations of the form [S . . . αi . . . [ω . . . [βi . . . ]]] are degraded, where α ∈ A, β ∈ B, ω ∈ X and α and β need to enter a relation.

Due to their selectivity, intervention effects are often assumed to arise because of clashing semantic/pragmatic properties of the dependencies under consideration and the harmful interveners that can be obviated. In the following sections I will discuss two kinds of intervention effects along these lines. Section . discusses negative island phenomena. Section . is devoted to Beck effects with negation. In each case, I will attempt to give a comprehensive empirical overview and point to specific semantic/pragmatic accounts of the phenomenon under consideration. In section . I will conclude by tying together the findings from the various sections.

.. N 

..................................................................................................................................

... Expressions subject to negative islands We saw that in English wh-interrogatives, one wh-phrase is fronted to the beginning of the clause in order to form an operator-variable dependency at logical form. This is true for all kinds of wh-expressions including wh-adverbials like how, why, and when, and wh-phrases asking for degrees such as how many cars and how fast.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 ()

  a. b. c. d. e. f. g.

What did John fix _? Which car did John fix _ How did John fix the car _? How many cars did John fix _? How fast did John fix the car _? Why did John fix the car _? When did John fix the car _?

With the exception of why, all wh-expressions from () that do not ask for individuals, however, are subject to intervention by negation. That is, an operator-variable dependency cannot be established across an intervening negation. This constraint is called the negative island constraint (Ross ; Rizzi ). ()

a. b. c. d. e. f. g.

What didn’t John fix _? Which car didn’t John fix _ *How didn’t John fix the car _? *How many cars didn’t John fix _? *How fast didn’t John fix the car _? Why didn’t John fix the car _? *When didn’t John fix the car _?

For independent reasons Rizzi () argues that why in cases like (f ) is actually not moved to but rather inserted in the clause-initial position (see also Ko ). No operatorvariable dependency needs to be established in such cases (Bromberger ). This explains why (f ) is acceptable. In fact, other negative expressions like no one, few boys, and only John cause similar problems for operator-variable dependencies (Rizzi ). () illustrates this for howinterrogatives: ()

a. b. c. d.

How did John fix the car _? *How did no one fix the car _? *How did few boys fix the car _? *How did only John fix the car _?

What these problematic interveners share is that they are entailment reversing. In particular, they are Strawson-downward monotonic (von Fintel ). This means that they reverse the entailment observed between two sentences, provided that their presuppositions are true.² To see this notice that (a) entails (b). However, in () and

² Strawson-upward monotonic expressions such as indefinites do not reverse entailment. Strawsonupward and -downward monotonicity are defined as follows, where ⇒ stands for entailment: (i) a. f ∈ D〈σ, τ〉 is Strawson-upward monotonic iff for all a, b ∈ Dσ, where a ⇒ b and f(b) presupposes r, f(a) ∧ r ⇒ f(b). (ii) b. f ∈ D〈σ, τ〉 is Strawson-downward monotonic iff for all a, b ∈ Dσ, where a ⇒ b and f(a) presupposes r, f(b) ∧ r ⇒ f(a).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



() the entailment is reversed. In () the tautologous presupposition is assumed to hold and in () the one contributed by only that John fixed the car quickly (Horn ). ()

a. John fixed the car quickly. b. John fixed the car.

()

a. No one / few boys fixed the car. b. No one / few boys fixed the car quickly.

()

a. John fixed the car quickly and only John fixed the car. b. Only John fixed the car quickly.

As expected, these Strawson-downward monotonic expressions do not cause an intervention effect with wh-expresssions asking for individuals: ()

a. Which car did no one fix _? b. Which car did few boys fix _? c. Which car did only John fix _?

While these facts are more or less accepted, it is sometimes stated that the degree to which fronting across interveners is possible varies with the fronted wh-expression (see e.g. Cinque ; Rizzi ; Rooryck ). I do not engage in this discussion here as the facts are not entirely settled. To be on the safe side, the examples are marked as equally unacceptable. Obtaining more controlled judgments would be useful here and thus constitutes an avenue for future research.³

... Why negative islands are not syntactic in nature Rizzi () proposed an influential syntactic account in terms of relativized minimality. The key intuition is as follows. An interrogative operator Qwh at the top of the clause and a wh-expression must undergo a feature relation followed by overt or covert movement of the latter to the specifier of the former. Now, a neg feature on a negative element intervening between the operator and a wh-expresssion like how specified for −individual will block the relation. Such a neg feature will, however, not block the relation if the wh-expression is specified for +individual, like who. Crucially on such a syntactic account, both strong and weak islands are an all-or-nothing phenomenon. That is, either one observes an effect of degradedness with a certain class of interveners or one does not. This seems to be incorrect, as I now show.

.... The influence of the semantics of the predicate Szabolcsi and Zwarts () show that dependencies formed with wh-expressions asking for individuals can be subject to weak islands as well, contrary to what we have ³ Negative islands are a subtype of so-called weak islands. Other subtypes are factive and wh-islands.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

seen so far. In particular, they are subject to intervention if there is a uniqueness implication present, as is necessary with the predicate to get this letter from x but not with the predicate to show this letter to x. A particular letter can only be from one individual, but it can be shown to multiple individuals. Now negative island effects are sensitive to the presence or absence of this implication. Whom can be fronted across negation in (b) without the uniqueness property, while it cannot be so in (b) with the uniqueness property present.⁴ ()

()

a. To whom did you show this letter _? b. To whom didn’t you show this letter _?

(Szabolcsi and Zwarts : (a))

a. From whom did you get this letter _? b. *From whom didn’t you get this letter _?

(Szabolcsi and Zwarts : (b))

Now, given the data reviewed, syntactic approaches to weak islands draw a strong distinction between wh-expressions asking for individuals and those asking for degrees, manners, time points, and the like. This is called into question by wh-expressions asking for individuals that are subject to weak islands. Moreover, it is not clear how a syntactic account could differentiate between the wh-expressions in, say, (b) and (b).

.... Contextual alleviation of negative islands The second reason for assuming that negative islands are semantic in nature is due to the fact that they can be alleviated by context. The observation of this fact goes back to Kroch (), the original version of which was written already in . For instance, in the context of A’s question in (), B’s utterance of what one would expect to cause a negative island effect is acceptable.⁵ () A: How much have beans been costing lately? B: The price has been jumping around so much, you’d do better to ask: How much haven’t they cost? (Kroch : ()) Apart from such ironic contexts, negative islands are alleviated when, intuitively speaking, the context restricts the domain of quantification for the wh-expression. Abrusán () makes this explicit by giving examples listing salient answers as in (), where (b) is based on an example in Kroch . ()

a. ?How didn’t John behave _: quickly or responsively? (Abrusán : , ()) b. Among the following, how many points did Iverson not score _? A:  B:  (Abrusán : , ())

⁴ To be sure, there might be a reading under which (12b) is acceptable as well, namely one where this letter is understood as referring to a type of letter rather than a particular token (Szabolcsi and Zwarts 1992). There can of course be multiple instances of a type. ⁵ In light of such contextual alleviation, it might be better to mark the degradedness of negative islands as # rather than as *. I follow the established practice and use * nevertheless.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



Arguably the effect seen in () extends to Kroch’s example in () with wh-movement away from an extraposed relative clause. Here it is the relative clause that serves to restrict the domain of quantification. ()

How much didn’t you pay that you were supposed to?

(Kroch : ())

To account for such contextual alleviation, syntactic accounts often rely on a distinction between referentiality and non-referentiality (e.g. Pesetsky ; Cinque ; Rizzi ). Note, however, that this would entail that wh-phrases like which NP are referential in a way that those such as how—without specific context—are not; only referential wh-phrases are hypothesized to be able to undergo a dependency across a negative island. Unfortunately, it is not clear at all what syntactic “referentiality” of wh-phrases could mean on the semantic side (Cresti ; Rullmann b). Wh-expressions are by their very definition non-referential.

.... Modal obviation phenomena A third argument for a semantic treatment of negative island effects comes from so-called modal obviation phenomena. Following Kuno and Takami (), Fox and Hackl () show that the addition of an existential or a universal modal can affect the availability of negative island effects in degree questions. In particular, having an existential modal under negation, (a), or a universal one above negation, (b), obviates the weak island effect otherwise obtained with negation. ()

a. How much radiation is the company not allowed to expose its workers to _? (Fox and Hackl : (a)) b. How much radiation is the company required not to expose its workers to _? (Fox and Hackl : (a))

Crucially, in each case the reversal of the relative order of modal and negation does not lead to an obviation effect: ()

a. *How much food is the company allowed not to give its workers _? (Fox and Hackl : (b)) b. *How much food is the company not required to give its workers _? (Fox and Hackl : (b))

Abrusán (, ) notes that the modal obviation pattern extends to manner-questions with negation as well. The attitude predicate hope is assumed to be a necessity modal following Hintikka (). ()

a. How is John not allowed to behave _ at the party? b. How do you hope that John did not to behave_?

(Abrusán : )

Arguably, modal obviation phenomena are even more problematic for syntactic accounts than contextual alleviation. The addition of a modal to a sentence is not expected to lead to a

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

change in the features associated with negation or the wh-expressions. If anything, additional material might be expected to add a further relativized minimality effect.

... Semantic accounts of negative islands The approaches to be discussed build on Szabolcsi and Zwarts’s () suggestion that negative island effects are due to a kind of maximality failure.

.... Rullmann (b) Since von Stechow () it has been standard to assume that natural language semantics makes use of degrees and that how-interrogatives such as () are questions for degrees. Rullmann (b) suggests that (a) asks for the maximal degree such that the addressee is tall to that degree. If the addressee is exactly  cm tall, then this is the relevant maximal degree the question is asking for. Similarly (b) asks for the maximal degree to which the addressee is not tall. In the situation just mentioned, the addressee is not tall to . cm, and also not to . cm, and so on. In fact, there is no maximal degree to which the addressee is not tall. But if there is no maximal degree to which the addressee is not tall, then asking the interrogative does not make much sense. This would be the reason for the degradedness of (b). () a. How tall are you _? b. *How tall aren’t you _? Beck and Rullmann () notice that interrogatives like () constitute a problem for this kind of reasoning. If, say, three eggs are sufficient for baking the cake, then four eggs will also be sufficient, and similarly so for any larger quantity. That is, there is no maximal number of eggs that is sufficient to bake the cake, in parallel to the situation in (b). Why then is () acceptable? () How many eggs are sufficient (to bake this cake)?

(Beck and Rullmann : ())

.... Fox and Hackl () As a response to this puzzle, Fox and Hackl suggest that measurement scales used in natural language are universally dense. That is, for any two degrees d and d0 , where d < d0 , there is another degree d00 , such that d < d00 < d0 . Moreover, they shift the problem from non-existent maximal degrees to non-existent maximal answers. For concreteness, assume as in () that tall is a predicate of degrees d yielding true for an individual x if and only if x is d-tall (see e.g. Heim , but cf. Kennedy  for a compatible alternative). ()

⟦tall⟧w = λdd.λxe.x is d-tall in w

In Karttunen’s  theory, wh-questions denote sets of proposition or rather characteristic functions thereof. Assuming now that how quantifies over degrees, the denotations of the

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



interrogatives in () are as in (), where Dd is the domain of degrees. The first denotation is the set of propositions saying that the addressee is d-tall, for some degree d, whereas the second one is the set of propositions saying that the addressee is not d-tall, for some degree d. ()

a. ⟦(a)⟧w,g = λp.∃d[d ∈ Dd ∧ p = λw. the addressee is d-tall in w] b. ⟦(a)⟧w,g = λp.∃d[d ∈ Dd ∧ p = λw. the addressee is not d-tall in w]

The propositions in the denotations can be thought of as potential mention-some answers to the question. Fox and Hackl now assume, following Dayal (), that the speaker of a question presupposes that there is a maximal answer to the question, where the maximal answer is defined as the unique true proposition in the question denotation entailing all other true propositions in the denotation. By uttering a question the speaker wishes to be told that maximal answer. So by (a), a speaker asking (a) wants to know the unique true proposition of the form that the addressee is tall to degree d entailing all other such true propositions. If the addressee is exactly  cm tall, then that proposition is that the addressee is tall to degree  cm. For (b), however, the situation is different. In the same situation, the addressee is not tall to degree . cm, and the same for any larger degree, as seen above. Is . cm the relevant degree then? No, because under the assumption that measurement scales are dense, there is always a smaller degree between  cm and . cm, call it d, such that the addressee is not tall to that degree. Moreover, there is another such degree between  cm and d. And so on. Thus the maximal answer to ⟦(a)⟧w,g is necessarily undefined, unlike the one to ⟦(b)⟧w,g. It is reasonable to assume that such questions are not used. This approach can also deal with modal obviation phenomena. Consider () from (b) again. Its denotation would be (). Assume the law requires that the workers are not exposed to d0 -amount of radiation. Then the proposition that the company is required not to expose its workers to d0 -amount of radiation is the unique proposition entailing all other true propositions in the question denotation. For any d00 > d0 , this proposition entails that the company is required not to expose its workers to d00 -amount of radiation. This accounts for the acceptability of (). Parallel reasoning extends to (a) above as negation scoping over an existential modal is equivalent to negation scoping below a universal modal. ()

How much radiation is the company required not to expose its workers to _? (Fox and Hackl : (a))

()

⟦()⟧w,g = λp.∃d[d ∈ Dd ∧ p = λw. in all worlds w0 accessible from w the company does not expose its workers to d-amount of radiation]

Crucially, the situation for () is parallel to the one just seen for (). Assume that exactly three eggs are sufficient for baking the cake. This entails that . eggs are sufficient, but also that . eggs are sufficient and so on. Thus the unique true proposition entailing all other true propositions in the question denotation is that  eggs are sufficient to bake the cake. Of course, this necessitates that dense scales are extended to cases where such scales are normally not assumed, as when counting eggs. Similarly, it is not clear what, say, . cars would amount to, as would be required in order to deal with ().

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 ()

  *How many cars didn’t John fix _?

For this reason, Fox and Hackl assume that there is a level of interpretation, the deductive system, which is blind to contextual information and common knowledge and works with dense scales for all kinds of counting. This necessitates a strong version of modularity (see also Fox , Gajewski  for the related notion of a deductive system), where context cannot affect the granularity of scales used by the deductive system. This raises the question how context should then be able to alleviate negative islands. Recall () from above. Fox and Hackl suggest that in such cases contextual restriction happens. That is, not is restricted by a covert variable containing only the numerals  and . Intersection of the set of dense numerals with this set yields a non-dense scale. In other words, context cannot affect the granularity of scales directly but can do so indirectly by semantic restriction. () Among the following, how many points did Iverson not score _? A:  B:  (Abrusán : , ()) Finally, it is not immediately clear how to extend the density-based account to negative islands in non-degree-questions such as () (but see Fox b). ()

*How didn’t John fix the car _?

.... Abrusán () Giving up universally dense scales, Abrusán retains Fox and Hackl’s suggestion that the speaker presupposition regarding a maximal answer is central to accounting for negative islands but combines it with certain assumptions about denotation domains harking back to Szabolcsi and Zwarts’s  proposal. For degree-questions such as those in (), Abrusán suggests that they actually ask for an interval of degrees, that is a set of degrees (Abrusán and Spector ; Schwarzschild and Wilkinson ). Accordingly, tall is a function from intervals I to functions from individuals x to truth values yielding true for x if and only if x’s height is a degree in I, as in (a). ()

⟦tall⟧w = λIdt.λxe.x’s height is in I in w

How must now quantify over intervals rather than degrees. With this the denotations for (a) and (b) are then as in (). ()

a. ⟦(a)⟧w,g = λp.∃I[I ∈ Ddt ∧ p = λw. the addressee’s height is in I in w] b. ⟦(b)⟧w,g = λp.∃I[I ∈ Ddt ∧ p = λw. the addressee’s height is not in I in w]

Given our assumptions from above, a speaker of (a) asks for the unique true proposition of the form that the addressee’s height is in interval I entailing all other such true propositions. If the addressee is exactly  cm that proposition is that the addressee’s height is in interval [ cm, cm], which is equivalent to saying that the addressee is exactly  cm tall. With (b), however, a problem arises again. There are at least two non-overlapping

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



intervals such that the addressee’s height is not in them: [, . cm] and [. cm, ∞]. Now, the proposition that the addressee’s height is in [, . cm] does not entail that the addressee’s height is not in [. cm, ∞] and vice versa. Completely parallel considerations apply to smaller intervals based on the two mentioned. Thus again there is no unique answer in the case of (b) in the sense required by the answer operator. Abrusán assumes that manner predicates like quickly are modifiers of eventualities. Therefore how in () quantifies over such modifiers in parallel to the quantification over intervals just seen. That is, the denotation of () is as in (). () ⟦⟧w,g = λp.∃m[m ∈ Dm ∧ p = λw. John did not fix the car in manner m in w] Crucially, Abrusán suggests that every set of manners must contain for each manner both its contrary and the middle between the two contraries, where two contraries contradict each other by definition. Assume that Dm in () is restricted contextually to manners of speed, then it contains quickly together with slowly and with regular speed. Assume now that John did not fix the car slowly. Would that John did not fix the car slowly be the unique true answer entailing all other true answers in ()? No, if John actually fixed the car quickly, that John did not fix the car with regular speed would also be true. But the latter proposition is not entailed by the former one. And vice versa if John fixed the car with regular speed. Then that John did not fix the car quickly would be true, which is also not entailed. Therefore no unique true answer can be found and () is degraded. While this improves over Fox and Hackl’s account, it remains to be seen how the assumption that sets of manners always contain contraries and middles can be substantiated. For now it is a stipulation. Abrusán’s proposal can also deal with modal obviation facts and contextual alleviation. The logic of the argument is similar to that seen with Fox and Hackl’s  proposal. I refer the reader to Abrusán () for details. I want to briefly return to discussion of the influence of the verbal predicate on negative islands. Abrusán suggests that given the punctual predicate die, (a) is degraded because there are infinitely many time points where Mary did not die. Moreover, the time points and the resulting propositions are not ordered by entailment. Thus no unique answer can be found in parallel to what we saw above. With the stative predicate in (b), on the other hand, maximal intervals where the addressee did not feel happy can be found, accounting for its acceptability. ()

a. *When did Mary not die? b. When didn’t you feel happy?

(Abrusán : , ())

But recall the data in () from above, which seem related to the ones in (). Given the logic of the argument, then (a) should be degraded because there are infinitely many individuals from whom the addressee did not get this letter. This is, however, not true given real world knowledge. There is a finite set of individuals, though maybe not listable. This suggests that even on Abrusán’s account one might need to resort to something like dense scales for entities other than degrees. That is, in order to account for () one might need dense scales of individuals. But this would bring the account closer to Fox and Hackl’s () one, and in the end Abrusán’s criticism of universal dense scales might not hold up.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

() a. *From whom didn’t you get this letter _? b. To whom didn’t you show this letter _?

(Szabolcsi and Zwarts : (b))

I now turn to discussion of Beck effects, showing that they are substantively different from negative islands.

.. B 

..................................................................................................................................

... The nature of Beck effects .... The unacceptability in Beck effects As noted by Rizzi () and Beck () a negative quantifier like niemand in German creates an intervention effect in (b) in contrast to the minimally different (a). ()

a. Wen hat Hans _ wo gesehen? whom has Hans where seen ‘Who did Hans see where?’ b. *Wen hat niemand _ wo gesehen? whom has nobody where seen

(Beck : (b))

What actually goes wrong in (b)? To see this one must notice that a multiple wh-interrogative such as (a) can either be answered by a single-pair answer—that is, an answer specifying the unique pair making the relation denoted by the verb true—or by a multiple-pair answer—that is, an answer specifying all pairs making that relation true. In other words, both (a) and (b) are possible answers to (a). ()

a. He saw Anton in the office. b. He saw Anton in the office, Maria in the cinema, and . . .

One might think, of course, that the single-pair answer is a special case of the multiple-pair answer. Beck effects, however, suggest otherwise. It is sometimes noted in the literature that an interrogative such as (b) is actually acceptable under a single-pair interpretation but not on a multiple-pair interpretation (e.g. Pesetsky ; Beck ; Kotek ). That is, it can be answered by A in () but not by A0 . This state of affairs would be unexpected if single- and multiple-pair answers were due to the same interpretation of the multiple whinterrogative. ()

a. No one saw Anton in the office. b. No one saw Anton in the office, no one saw Maria in the cinema, and . . .

In other words, (b) is judged degraded on its multiple but not on its single-pair interpretation. The latter is more restricted (Dayal , ). For instance, its prosody does not correspond to the one of regular wh-questions including those with multiple-pair

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



interpretations (see Pope ). This is why we judge (b) degraded when presented on its own. Beck () moreover notes that embedding a multiple wh-interrogative with an intervener under predicates like enumerate blocks the single-pair interpretation. The degradedness then comes out clearly, as the contrast in () inspired by Beck’s () shows. ()

*Luise zählt auf, welchen Professor niemand _ wohin eingeladen hat. Luise enumerates which professor nobody where.to invited has

.... Beck effects are about wh-in-situ expressions and not weak islands This raises the question whether (b) is simply an instance of a negative island as discussed above. For instance, one might think that negative islands for questions over individuals only become visible once we look at multiple wh-interrogatives. That is, maybe negative islands for questions over individuals manifest themselves by blocking the multiple-pair interpretation thereby disambiguating the questions. The fact that (b) significantly improves if the non-fronted wh-expression is placed before the intervening negative quantifier, as in (), strongly suggests otherwise. That is, () can have the multiple-pair interpretation and is thus answerable by (b). () Wen hat wo niemand _ gesehen? who has where nobody seen ‘For which individual x and place y, no one saw x at y?’ That we are not dealing with an instance of a negative island is also suggested by the fact that unlike with the latter modal obviation does not seem to exist. For instance, possibility modals below negation do not obviate Beck effects, as the paradigm in () shows. zu treffen? () a. *Wen ist niemandem erlaubt worden _ wo who is nobody allowed become where to meet b. Wen ist wo niemandem erlaubt worden _ zu treffen? who is where nobody allowed become to meet ‘Which person x and place y are such that no one was allowed to meet x at y?’ Returning to (), its acceptability suggests that it is a relationship between the non-fronted wh-expression and some higher element that is prohibited. This is unlike what we know from negative islands. The generalization that Beck effects are about wh-in-situ expressions is also supported by cross-linguistic considerations. For instance, wh-in-situ interrogatives in Japanese such as in () show behavior comparable to that of German multiple wh-interrogatives (Hoji ; Beck and Kim ; Hagstrom ; Kim ; Tomioka ). An NPI like anyone causes intervention when c-commanding a wh-in-situ expression but not when the latter ccommands the former. Note that here issues about single vs. multiple-pair interpretations do not even arise. Korean shows a completely parallel situation.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

yom-ana-katta-no () a. ?*Daremo nani-o anyone what- read--- b. Nani-o daremo _ yom-ana-katta-no what- anyone read--- ‘What did no one read?’

(Tomioka : (a))

A prevalent intuition in the literature is that languages like German, Japanese, and Korean show Beck effects because they are so-called scrambling languages. In such languages the mutual scope between two expressions is largely determined by surface order. The degraded interrogatives seen above are degraded then because the scope of the wh-in-situ expression is restricted to its surface position, while for semantic reasons it should take scope above the intervener. Kotek () shows that this intuition is also supported by facts from English. At first sight, it seems as if English did not show Beck effects given that Q in () is acceptable in contrast to the German example above. That is, the English multiple wh-interrogative is acceptable both on a single- and a multiple-pair interpretation. ()

Q: Who did no one see where? A: No one saw Tony in the office. A0 : No one saw Tony in the office, no one saw Mary in the cinema, . . .

Pesetsky () shows that intervention effects exist in English as well. In particular, in multiple which-interrogatives the so-called superiority condition can be violated. That is, here it is not always the structurally highest wh-phrase that must be fronted, as (a) shows. Now, in such violations of superiority Beck effects emerge, as the degradedness of (c) relative to (b) shows. Kotek argues that this difference is due to the fact that in (c) but not in (b) the scope of the wh-in-situ expression is restricted to its surface position. ()

a. Which student did Mary give which book to _? b. Which book did no one give _ to which student? c. ?*Which student did no one give which book to _?

(Pesetsky : ())

In general, with regards to which languages show the kind of intervention effect under discussion, Beck (: ) lists the following: Dutch, English, German, French, Hindi/ Urdu, Japanese, Korean, Malayalam, Mandarin, Passamaquoddy, Persian, Thai, and Turkish.

.... Problematic interveners and affected wh-expressions Apart from negative quantifiers, what is the class of problematic interveners with regards to Beck effects? I cannot give an exhaustive list here for reasons of space. But I want to give a few examples at least. See also the following sections. As probably expected, negation in general causes intervention. The same holds for other languages. In fact, all kinds of Strawson-downward monotonic expressions are interveners

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



(Beck ; Honcoop ; Mayr ). But in addition the focus operator even is also a problematic intervener, even though it is not Strawson-downward monotonic. () a. Which girl did only Mary introduce _ to which boy? b. ?*Which boy did only Mary introduce which girl to _?

(Pesetsky : ())

() Wen hat wann sogar der Hans _ angerufen? who has when even the Hans called ‘Who did even Hans call when?’

(Mayr : ())

Universal quantifiers are also problematic interveners. A superiority obeying multiple whinterrogative such as Q in () can be answered by both A and A0 . Crucially, A corresponds to an interpretation of Q where the universal quantifier is scoped out of the interrogative (Groenendijk and Stokhof ; Engdahl ; Chierchia ). On that interpretation it can be paraphrased as ‘For every x, tell me which newspaper x wrote to about which book?’. With A0 , however, the universal quantifier is interpreted in its surface scope position, as is also apparent from the fact that it figures in the multiple-pair answer repeatedly. Now, the superiority violating Q0 can only be answered by A. That is, it only seems to have an interpretation where the universal quantifier is scoped out. The reason for this would be that on the other interpretation the universal would cause an intervention effect. This is also true for German (Beck ). ()

Q: Which newspaper did everyone write to _ about which book? Q0 : Which book did everyone write to which newspaper about _? A: Bill wrote to the New York Times about book X, Mary wrote to the Boston Globe about book Y, and Tom wrote to the Maquoketa Sentinel about book Z. A0 : Everyone wrote to the New York Times about book X, everyone wrote to the Boston Globe about book Y, and everyone wrote to the Maquoketa Sentinel about book Z. (Pesetsky : f.)

From this cursory look it seems that it is quantificational expressions or scope bearing expressions that cause Beck effects. But this generalization has been called into question, as we will see once we turn to existing accounts. Which dependencies with in-situ wh-expressions are affected negatively by interveners? The examples so far have shown that wh-expressions asking for individuals, places, and times are affected. In addition, expressions asking for degrees and manners are also affected as () and () show: () a. Wen hat wie niemand _ getroffen? who has how nobody seen ‘Which x and which manner y are such that no one met x in y?’ b. *Wen hat niemand _ wie getroffen? who has nobody how seen

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

() a. Wen hat wie oft niemand _ getroffen? who has how often nobody seen ‘Which x and which degree d are such that no one met x to d?’ b. *Wen hat niemand _ wie oft getroffen? who has nobody how often seen This differentiates Beck effects strongly from weak islands. They are unselective with regards to the kinds of wh-dependencies that are affected. This suggests that the explanation for the former does not lie in the differing natures of the bound variables as we have seen for the latter.⁶

... Existing accounts of Beck effects .... Intervention by focus Beck () argues that cross-linguistically it is focus operators that cause degradedness more reliably than simple quantifiers such as no one, everyone, and so on. In this she follows Kim (). She thus draws the conclusion that it is not the quantificational nature of the interveners seen so far that cause degradedness but rather the fact that they associate with focus (see also Beck and Kim ; Cable ; Kotek ). Consider the Korean case of intervention by only in (b) and the acceptable (a) without intervention. Assume the LFs in () using English words for simplicity. () a. Suna-ka muôs-ûl sa-ss-ni? Suna- what- buy-- ‘What did Suna buy?’ b. *Minsu-man nuku-lûl manna-ss-ni? Minsu-only who- meet--

(Kim : (a)) (Kim : (a))

() a. [CP  [IP Suna bought what ]] b. [CP0  [IP” onlyAlt [IP0 MinsuF meet who ]]] Without going into too much detail, Beck’s proposal works as follows. Following Rooth (, ) each expression has both an ordinary and a focus semantic value, where the former is what we are used to and the latter is a set. Now according to Beck, wh-expressions do not have a defined ordinary semantic value. As their focus value, they contribute alternatives, that is, sets of individuals which combine point-wise with the rest of the sentence. This means that IP in (a) does not have a defined ordinary value either but only a defined focus value corresponding to a set of propositions varying in the individual alternatives contributed by the wh-expression. Now, the question operator Q takes the focus value of IP and makes it the ordinary value of CP. This way the undefinedness of the

⁶ Moreover, so-called wh-scope marking constructions are also affected. See Beck ().

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



ordinary value of IP does not have an effect. The end result is a Karttunen-denotation for (a): () a. ⟦IP⟧o,w = undefined b. ⟦IP⟧f,w = {p : λw0 .∃x[ Suna bought x in w0 ]} = ⟦CP⟧o,w For (b), however, things work a bit differently. The focus association operator only is an unselective binder. That is, it makes use of all the focus alternatives in its scope; those contributed by the F-mark on Minsu and those contributed by the wh-in-situ expression. It requires that the ordinary value of its sister IP0 be true. But it is undefined, as before. Thus the ordinary value of IP00 is undefined as well. Now only makes the singleton containing the ordinary value of IP00 the focus value of IP00 , which is also undefined. The Q-operator finally takes this and makes it the ordinary value of CP. That is, the CP has no defined value: () a. (i) ⟦IP0 ⟧o,w = undefined (ii) ⟦IP0 ⟧ f,w = {p : ∃x.∃y[p = λw0 .y met x in w0 ]} b. (i) ⟦IP0 ⟧o,w = defined only if ⟦IP0 ⟧o,w =  among other things = undefined (ii) ⟦IP00 ⟧f,w= {⟦IP00 ⟧o,w} = undefined = ⟦CP0 ⟧o,w If the wh-expression in (b) were to be fronted before the intervener, a defined question denotation could be derived. This is so because the undefinedness of the ordinary value of the wh-expression would now be contributed after the application of only. Therefore the focus value would remain defined. That is, with fronting of the wh-in-situ across the intervener, the composition works as the one for (a). This account can be straightforwardly extended to multiple wh-interrogatives.

.... Intervention by scope bearing expressions So how then would, say, a negative quantifier cause intervention? It is well known that the truth conditions of sentences containing, for instance, quantifiers can be altered by focus (Rooth ). However, it is also known that quantifiers need not associate with focus (see Beaver and Clark ,  a.o.). It is for this reason that Beck assumes it is actually a silent operator ~ that evaluates focus (Rooth ). Operators like only must make use of ~, and quantifiers in German and English must do so, too. In languages like Korean ~ need not always be present with no one, whereby it does not reliably cause intervention. At this point it becomes crucial to understand why ~ seems to be cross-linguistically necessarily present with only. Such considerations highlight a problematic feature of Beck’s account. While it is technically elegant, its predictive power as to what counts as a problematic intervener and what does not is limited. It has been noted in the literature, however, that potentially a different empirical

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

generalization can be drawn, whereby problematic interveners share a feature other than association with focus with each other (Honcoop ; Haida ; Mayr ). Mayr () It is argued that at least in German the class of problematic interveners is fully predictable. Mayr suggests that only those expressions that do not commute scopally with existential quantifiers cause a Beck effect. An expression α commutes scopally with an existential quantifier if and only if ∃x.α.ϕ = α.∃x.ϕ, that is, if the mutual scope of α and the existential quantifier does not matter for the truth conditions. Mayr argues that this is the reason why indefinites or existential quantifiers never cause intervention unlike negative expressions. The following pair illustrates: ()

a. Wen haben mehr als drei Studenten _ wann eingeladen? who have more than three students when invited ‘Who did more than three students invite when?’ b. *Wen haben weniger als drei Studenten _ wann eingeladen? who have fewer than three students when invited (Mayr : ())

Simplifying considerably, assume that positive indefinites denote existential quantifiers themselves and negative indefinites denote negative quantifiers. Then it is plain why the former do not cause intervention whereas the latter do: ∃x.∃y.ϕ = ∃y.∃x.ϕ, but ¬∃x.∃y.ϕ 6¼ ∃y.¬∃x.ϕ. Now, Mayr shows that all of the problematic interveners seen so far do not scopally commute with existential quantifiers. That is, the focus operators only and even are also predicted to cause intervention. As discussed by Haida and Repp () additive focus operators like also would be crucial to distinguish between Beck’s proposal and Mayr’s. The reason is that they contribute a simple existential presupposition and thus would be classified as scopally commutative. Their experimental data actually go in this direction. That is, the German version of also, unlike the other focus operators, does not seem to cause intervention effects, which would not be predicted by Beck. Why would scopal commutativity be relevant though? Mayr suggests that the existential presupposition contributed by a wh-expression must match the space of possible answers derived from the question denotation. Now, with commutative operators we get such a match but not with non-commutative ones, which is reflected by the (non-)equivalences referred to above. Honcoop (), Haida () The authors draw attention to the fact that intervention in anaphoric dependencies empirically looks somewhat similar to Beck effects. In particular, it is well known that indefinites can be anaphorically related to pronouns they do not c-command—so-called donkey anaphora—as shown in (a) and indicated by co-indexing. Crucially, replacing the higher indefinite with a universal or negative quantifier as in (b) and (c) leads to unacceptability (see Heim  a.m.o.). The quantifiers make the indefinite inaccessible for entering into an anaphoric relation with the non-c-commanded pronoun. ()

a. If a farmerj owns a donkeyi, hej beats iti. b. *If every farmer owns a donkeyi, he beats iti. c. *If no farmer owns a donkeyi, he beats iti.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



Honcoop and Haida now suggest that the problematic interveners causing Beck effects should be seen as establishing inaccessible domains for anaphora similar to what is happening in ().⁷ This makes sense under the assumption that the wh-in-situ expression in (), repeated from above, must undergo an anaphoric relationship with an operator at the top of the clause. That is, assume an LF like () and assume that the question operator must bind the variable introduced by the index  on which book. ()

?*Which student did no one give which book to _?

()

[ ? [ which student [ no one gave which book2 to t1 ]]]

(Pesetsky : ())

If the quantifier binds the variable contributed by which book unselectively, however, the question operator cannot bind it anymore. Thus no question denotation obtains, or at least not the intended one. Note that for this to work, the quantifier must not bind the variable corresponding to the trace of the fronted wh-expression. Otherwise, simple whinterrogatives with negative quantifiers would be equally unacceptable. That is, binding by no one must be somewhat selective.⁸ The class of problematic interveners predicted by Honcoop and Haida is close to what we have seen in the discussion of Mayr (). It should, however, be noted that focus operators like only do not seem to cause intervention in donkey anaphora, as in (), unlike what we have seen for Beck effects. This is somewhat unexpected on Honcoop’s and Haida’s accounts. () If only John has a donkeyi, he beats iti.

.... Intervention by anti-topics Tomioka notes that in a wh-interrogative in Japanese without a problematic intervener, the expression that occurs in the position where the problematic intervener would otherwise occur is typically marked as a topic, that is, is marked by wa in (). The same is shown to be true for Korean. () John -wa / ?-ga nani -o yon -da -no? J -/ - what- read-- ‘What did John read?’

(Tomioka : (a))

Now, Tomioka notices that the class of problematic interveners such as daremo in (a) cannot be marked with wa. That is, they cannot function as topics. They are anti-topics. Tomioka consequently draws on the observation that the material in a wh-interrogative that is not wh-marked must be old information (Prince ; Schwarzschild ). In

⁷ Honcoop () actually also intends his account for weak islands. ⁸ This essentially replicates the DRT approach to donkey anaphora as laid out in ch.  of Heim (). Note that this brings out a similarity to Beck’s () account. She applies the DRT strategy to the focus domain rather than the ordinary denotation.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

particular, daremo in (a) should be old information. However, it cannot be so because it is an anti-topic. On Tomioka’s account Beck effects are thus the consequence of two incompatible pragmatic constraints; one a constraint on non-wh-marked expressions occurring in a wh-interrogative and one a constraint on certain lexical expressions such as daremo. For (b) it is shown that the fronting of the wh-expression leads to deaccenting of the remaining material including daremo. Via this de-accenting, material that can otherwise not become a topic does do so, making (b) acceptable. While an intriguing idea, it is less clear whether the approach carries over to other languages. In particular, Mayr () argues that the correlation noted by Tomioka for Japanese and Korean does not hold up for German (but see Grohmann  for a different view).

.. C

.................................................................................................................................. I discussed two kinds of intervention effects caused by negation: negative islands and Beck effects. In both cases we have grounds to believe that the resulting intervention effect is semantic in nature. The main reasons for this are that negative islands, on the one hand, can be alleviated in various ways and that Beck effects, on the other hand, seem to be about a dependency that is closer to anaphora than syntactic movement. I therefore suggested that a unified approach is not called for. We saw that numerous issues remain open.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ......................................................................................................................

    , ,    ......................................................................................................................

 

.. I

.................................................................................................................................. I clauses often function as neutral requests for information, as in (), but they do not always do so. This chapter examines three types of interrogatives that, besides raising the issue posited by the interrogative, convey epistemic bias on the speaker’s side. In all three types, polarity—and in particular negation and negation-dependent items— plays a crucial role. () (I’m really curious:) Did Al help you? The first type of interrogative, to be examined in section ., is biased information-seeking questions like (), where negation places a central role in triggering the bias. Tag questions like (), examined in section ., constitute the second type. Here the relation between the polarity of the prejacent and the tag interrogative (reverse polarity vs. constant polarity) together with other surface cues bring crucial pragmatic distinctions. Section . turns to the third and final type: rhetorical questions, for example (), where the presence of strong negative polarity items like lift a finger mandatorily leads to rhetoricity and to a negative (intended) answer. ()

Didn’t Al help you?

()

Al helped you, didn’t he?

()

Did Al lift a finger to help you?

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

.. B - 

..................................................................................................................................

... Basic data The interrogative polar question forms in ()–() all raise, intuitively, the same issue: which proposition in the set {p, ¬p} is true, where p equals the proposition ‘that Amy arrived’ in these examples. Indeed, classical semantic analyses will assign the three forms the same truth-conditional content (Karttunen ; Groenendijk and Stokhof ). But the forms cannot be used interchangeably, witness (): () Did Amy arrive?

Positive Question (PosQ)

() Did Amy not arrive?

Low Negation Question (LowNQ)

() Didn’t Amy arrive?

High Negation Question (HiNQ)

() Context: Unbiased U.S. immigration officer asking a traveller on the citizen line. a. Are you a U.S. citizen? b. # Are you not a U.S. citizen? c. # Aren’t you a U.S. citizen? This means that, while the three forms may have the same truth-conditional content, they certainly do not have the same use-conditions. This is because each form has a characteristic profile in terms of two biases discussed in the literature: (a) original speaker bias (Ladd ; Romero and Han ) and (b) contextual evidence bias (Büring and Gunlogson ). We start with original speaker bias, which concerns the original expectation of the speaker with respect to proposition p, as defined in (). Positive questions (PosQs) can be used in a neutral way and thus convey no such bias, as we saw in (). As for its negative counterparts, an English High Negation question (HiNQ) of the form n’t p? mandatorily expresses original speaker bias for p (Ladd ), whereas the corresponding Low Negation question (LowNQ) not p? can be used in certain contexts without such bias (Romero and Han ). This is shown in (): In a context where the speaker has no prior expectation about p (= ‘that John drinks’), the LowNQs in (S) can be used but the HiNQ (S0 ) would be inappropriate: ()

Original speaker bias for a proposition p (Domaneschi, Romero, and Braun ; cf. Ladd : ) Belief or expectation of the speaker that p is true, based on his epistemic state prior to the current situational context and conversational exchange.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

, ,    ()



Context: The speaker is organizing a party and she is in charge of supplying the non-alcoholic beverages for teetotalers. The speaker is going through the list of invited people. The speaker has no previous belief or expectation about their drinking habits. A: Jane and Mary do not drink. S: OK. What about John? Does he not drink (either)? (Romero and Han ) S0 : # OK. What about John? Doesn’t he drink (either)?

While expressing original speaker bias, HiNQs have been argued to allow for two different interpretations (Ladd ). Under the first interpretation, labeled the “outer negation” reading and illustrated in (), the speaker double-checks p (= ‘that there is a good restaurant nearby’). Under the second reading, labeled the “inner negation” reading and shown in (), the speaker intuitively double-checks ¬p (= ‘that there isn’t a good restaurant nearby’). These two interpretations are disambiguated by the presence of Positive Polarity Items (PPIs) like already, too, and someone—which only allow for the outer reading, as in (a)—vs. Negative Polarity Items (NPIs) like either, yet, and anyone—which enforce the inner reading, as in (b). ()

A: You guys must be starving. You want to get something to eat? S: Yeah, isn’t there a vegetarian restaurant around here?

Outer-HiNQ

()

S: I’d like to take you guys out to dinner while I’m here—we’d have time to go somewhere around here before the evening session tonight, don’t you think? A: I guess, but there’s not really any place to go in Hyde Park. S: Oh, really, isn’t there a vegetarian restaurant around here? Inner-HiNQ

()

a. Did Amy arrive already / too / with someone? ✓Outer * Inner b. Didn’t Amy arrive yet / either / with anyone? * Outer ✓Inner

Romero and Han () add to the picture a fourth question form: Positive questions with epistemic really (really-PosQ) mandatorily convey original bias against p (i.e. for ¬p). The same bias is sometimes expressed with focal accent on the tensed verb.¹ This makes these forms infelicitous when the speaker is unbiased, as in (): ()

Unbiased immigration officer to traveller when entering the U.S.: a. # Are you really a U.S. citizen? really-PosQ b. # ARE you a U.S. citizen?

¹ Focus on the inflected verb has other uses as well, e.g. focus-marking on tense in (ia) or on the polatiry in (ib). The use the authors refer to intuitively emphasizes or insists on the truth of the proposition, as in (ii), and was christened “Verum Focus” by Höhle (): (i) a. Amy isn’t sick. She WAS sick. b. Grant isn’t here. Carly IS. (ii) I (really) AM tired.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

These observations are summarized in Table ..

Table .. Original speaker bias across forms according to Ladd () and Romero and Han () Original speaker bias

PosQ p?

LowNQ Not p?

Inner HiNQ n’t p + NPI?

Outer HiNQ n’t p + PPI?

really-PosQ really p?

mandatory bias for p

*

*

✓

✓

*

no mandatory bias

✓

✓

*

*

*

mandatory bias against p

*

*

*

*

✓

We turn to contextual evidence bias (Büring and Gunlogson ), defined in (). Its effects are illustrated in (). Büring and Gunlogson’s empirical generalizations are: (i) a PosQ p? is incompatible with contextual evidence bias against p, (ii) LowNQ requires contextual evidence against p,² and (iii) an outer HiNQ with a PPI is incompatible with contextual evidence for p. This is summarized in Table .. ()

Contextual evidence bias for a proposition p (Büring and Gunlogson : ) Expectation that p is true (NB: possibly contradicting prior belief of the speaker) induced by evidence that has just become mutually available to the participants in the current discourse situation.

()

Scenario: A enters S0 windowless computer room wearing a dripping wet raincoat (contextual evidence for p=‘it is raining’). a. S: Is it raining? b. S: # Is it sunny?

Table .. Contextual evidence bias across forms according to Büring and Gunlogson’s (). Contextual evidence

PosQ p?

LowNQ Not p?

Outer-HiNQ n’t p + PPI?

cont. evidence for p

✓

*

*

no cont. evidence

✓

*

✓

cont. evidence against p

*

✓

✓

After these pioneering works, authors have assumed a partial and at times disagreeing mixture of the characterizations above, with two main points of divergence. ² Inspired by Ladd’s inner/outer distinction, Büring and Gunlogson () label the second group as “inner negation” questions, but the actual examples they use correspond to LowNQs and not to HiNQs with NPIs.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

, ,   



First, in terms of kinds of bias, while most authors have concentrated on one type of bias and ignored the other, some authors have actively argued for a conflation of the two biases under one unified (temporally sensitive) notion (van Rooy and Šafářová ; cf. Northrup ) and some for treating them as two separate coordinates (AnderBois ; Sudo ). Second, authors disagree on the classification of negative question forms. LowNQs, inner-HiNQs, and outer-HiNQs are treated as one joint grammatical type in van Rooy and Šafářová (). Other authors group LowNQs and inner HiNQs together as having the same bias import while distinguishing them from outer HiNQs (Krifka ; Asher and Reese ). Yet other authors vindicate the bias difference between LowNQs and HiNQs, but deny that Ladd’s interpretations in ()–() constitute genuine ambiguity, thus classifying inner and outer HiNQs as one single grammatical type (AnderBois ; Northrup ; Goodhue ). Finally, LowNQs, inner-HiNQs, and outer-HiNQs have been treated as having each a separate bias profile (Romero and Han ; Reese ; Walkow ; Sudo ).

... Experimental studies These two empirical points of divergence have been in the spotlight of recent experimental studies. The first point—whether original speaker bias and contextual evidence bias should be conflated or should be kept as separate parameters—has been investigated for the evaluation (comprehension) and selection (production) of different polar question forms. In a series of comprehension studies, Roelofsen, Venhuizen, and Weidman Sassoon () investigate how natural the three question forms PosQ, LowNQ, and HiNQ are judged in each possible crossing of original speaker bias (OB) and contextual evidence bias (CE). Leaving aside the cells [OB:p/CE:p] and [OB:¬p/CE:¬p]³ as well as some cells that had experimental design problems,⁴ the average ratings obtained within each cell with a -point scale (=completely natural, =completely unnatural) are summarized in Table .. These partial results indicate some interaction between the two kinds of bias. For example, while PosQs are judged more natural with neutral contextual evidence (, in cell [OB:p/CE:neutral]) than with contextual evidence for ¬p (, in cell [OB:p/CE:¬p]), original bias also has an impact on them, as the decrease in naturalness in cell [OB:¬p/CE: neutral] (,) in comparison to the cell [OB:p/CE:neutral] (,) indicates. Domaneschi, Romero, and Braun () run a production study in which, for each of the relevant bias combinations,⁵ subjects have to choose and produce a PosQ, a LowNQ, a

³ Since having an original bias for p and then receiving contextual evidence for p typically does not prompt the speaker to ask a question, the cell [OB:p/CE:p] will not be considered. Similarly for the cell [OB:¬p/CE:¬p]. ⁴ As the authors themselves note, the positive contextual evidence provided in the materials was apparently too strong, rendering all question forms inappropriate in the cells [OB:neutral/CE:p] and [OB:¬p/CE:p]. Domaneschi, Romero, and Braun () consider the materials for cell [OB:neutral/ CE:¬p] problematic too. ⁵ Domaneschi, Romero, and Braun () leave the cell [OB:¬p/CE:neutral] out of the experiment arguing that the polar question form expected would be more complex than the other four basic forms tested.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

Table .. Averaged ratings per cell from Roelofsen, Venhuizen, and Weidman Sassoon () ORIGINAL BIAS EVIDENCE BIAS

p

Neutral

¬p

p Neutral

PosQ: , LowNQ: , HiNQ: ,

¬p

PosQ: , LowNQ: , HiNQ: ,

PosQ: , LowNQ: , HiNQ: ,

PosQ: , LowNQ: , HiNQ: ,

HiNQ, or a really-PosQ (or some other form if these are felt inappropriate). The preferred choices obtained for each cell are shown in Table .. These results clearly show that the question form chosen does not only depend on original speaker bias—since, if it did, the selected form would remain constant through each column— nor only on contextual evidence bias—since, if it did, the selected form would remain constant through each row. Instead, what determines the preferred question form is the crossing of

Table .. Results for preferred polar question form in Domaneschi, Romero, and Braun () ORIGINAL BIAS CONTEXTUAL EVIDENCE

p p

Neutral PosQ / really-PosQ

Neutral

HiNQ

PosQ

¬p

HiNQ

LowNQ

¬p really-PosQ

original and contextual evidence bias, hence supporting the fundamental distinction between the two bias kinds and the need to theoretically model both. The second issue concerns the classification of the different negative question forms— LowNQs, inner-HiNQs, and outer-HiNQ—into pragmatic bias types. That LowNQs and HiNQs constitute separate pragmatic types, contra van Rooy and Šafářová () and Krifka () a.o., is already shown by the results from Domaneschi et al. (), since the two forms are selected in different cells. As for HiNQs with NPIs vs. PPI and Ladd’s ambiguity, two sets of studies provide some tentative evidence, albeit in opposite directions. On the one hand, studies in Hartung () and Sailor () show that either-HiNQs are degraded in American English. The status of either-HiNQs is important because, while

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

, ,   



weak NPIs like any and yet are licensed in run-of-the-mill questions regardless of negation, strict NPIs like either are typically infelicitous without negation, witness ().⁶ Hence, for those speakers that judge either-HiNQs as unacceptable, an analysis where inner and outer HiNQs constitute one single grammatical type and the difference just lies in the pragmatics on NPIs, as in AnderBois () a.o., seems appropriate. ()

Have you eaten anything / yet / #either?

On the other hand, Romero et al. () present a study in which, starting with the speaker’s original bias for ¬p followed by evidence bias for p, the speaker is faced with two sorts of contexts: (a) contexts favoring Ladd’s outer reading and (b) contexts priming Ladd’s perceived inner interpretation. If HiNQs unambiguously express the outer reading, both contexts are predicted to lead to the same realization. If, however, HiNQs are genuinely ambiguous, a different realization is expected for each context type. The experimental results show an asymmetric realization pattern, thus lending tentative support for the genuine ambiguity treatment, as in for example Romero and Han ().

... Theoretical approaches Based on the somewhat fragmented empirical picture presented in .., several competing theoretical analyses have emerged. They can be roughly classified into three main lines: (A) expressed proposition line, (B)  line and (C) speech act line. Though some ideas are converging, there is no agreement or synthesis in the literature at this point, so we will present each line and briefly assess its main advantages and disadvantages. The key feature in the expressed proposition line is the special status assigned to the proposition in the sentence radical of the interrogative clause. In the PosQ (), the proposition in the sentence radical is ‘(that) Amy arrived’, whereas, in the LowNQ (), the proposition expressed by the sentence radical is ‘(that) Amy did not arrive’. The idea is that, even though both interrogative clauses induce the same partition {‘that Amy arrived’, ‘that Amy did not arrive’} of the Common Ground, the speaker chooses to express in the sentence radical whichever one of the two propositions is more “useful” to her. This notion of “usefulness” has been implemented in different ways. Van Rooy and Šafářová () couch it in terms of utility value within Decision Theory, roughly defined as in (): ()

A proposition p has a high utility value if: i. its becoming true brings the speaker closer to her goal or desires, or ii. its addition to the speaker’s belief state would trigger a wide revision of it.

⁶ But see Rullmann () for either in PosQs like (i), leading to a rhetorical flavor: (i) Nixon’s not very bright, but does Agnew have any brains either?

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

This means that the speaker may choose to express a (positive or negative) proposition q and thus utter q? because q is desirable—via (i)—or because q is unexpected/surprising and thus highly informative—via (ii). The first case is illustrated with requests like () (see Bolinger ) and could be extended to (S): ()

Can you help me?

The second case is exemplified by the negative question forms in (), all of which are treated by the authors as equally expressing the negative proposition ‘that Jane is not coming’ in the sentence radical. For the LowNQ (a) and the inner reading of HiNQ (b), the speaker originally expects p (= ‘Jane is coming’) but receives contextual evidence for ¬p (= ‘Jane is not coming’), and thus ¬p counts as surprising at a previous belief state, that is, at the speaker’s belief state before grounding the new evidence. For the outer reading of HiNQ (b), the same epistemic conflict arises, except that now ¬p counts as surprising (even) at the speaker’s current belief state. This means that, in van Rooy and Šafářová’s analysis, inner and outer high negation HiNQs are grammatically the same and that Ladd’s intuitive interpretations are pragmatically derived by anchoring the utility value to temporally different belief-desire states. ()

A: We are all here. Let’s go! a. S: Is Jane not coming? b. S: Isn’t Jane coming (either/too)?

A definite advantage of a decision-theoretical approach is that the utility value of a proposition is automatically calculated on the belief-desire state of the speaker. From this, it follows that biases may not always be epistemic but sometimes bouletic (or deontic) in nature, as suggested by data in Huddleston and Pullum (). The main shortcoming of the approach concerns the correlation between inner vs. outer HiNQs and the licensing of NPIs vs. PPIs. It is not worked out how the temporal anchoring of the utility value can predict the distribution of these items. A similar approach is taken by AnderBois (), who models the “usefulness” of the sentence radical proposition within the framework of Inquisitive Semantics (Ciardelli et al. ). AnderBois’s idea is that a PosQ p?, a LowNQ not p? and a HiNQ n’t p? share the same truth-conditional representation but differ crucially in their so-called ‘Projected Issues’, that is, on the issues that the speaker signals she is interested in for the immediately subsequent discourse. In (a,b), the presence of a proposition within the Projected Issues leads to evidence bias or bouletic bias for that proposition, while the absence of Projected Issues in (c) emphasizes the interest in what truth value p has and leads to original bias for p. As for Ladd’s ambiguity, AnderBois argues that HiNQs unambiguously have the outer reading and that Ladd’s intuited difference stems from what proposition the speaker is currently leaning towards. ()

Projected Issues according to AnderBois (): a. PosQ: Did Amy arrive? {‘Amy arrived’} b. LowNQ: Did Amy notlow arrive? {‘Amy did not arrive’} c. HiNQs: Didn’thigh Amy arrive? {}

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

, ,   



The main advantage of AnderBois’ approach is that it provides a unified account of both kinds of bias—original bias and evidence bias—while capturing their empirical differences. Its most notable limitation at this point is that it does not spell out how different types of NPIs pragmatically induce the effect of moving away from the speaker’s original belief. We turn to the V line, which is mostly concerned with original bias. Its keystone is the so-called  operator (Höhle ). In Romero and Han’s (, ) approach,  is defined as the metaconversational operator (). V is conveyed by the particle really and/or focus stress on the inflected verb in really-PosQs like (), and by high negation in HiNQs like () and ()–(). The presence of , together with the economy principle in (), secures original speaker bias in these question types. PosQs and LowNQs (with normal prosody) lack  and thus do not convey original bias. ()

λp.λws. 8w0 ∈ Epix(w) [ 8w00 ∈ Convx(w0 ) [ p ∈ Common-Groundw00 ] ] [Rough paraphrase: ‘x is sure that the proposition p should be added to the Common Ground’.]

()

Principle of Economy: Do not use a meta-conversational move unless necessary (to resolve epistemic conflict or to ensure Quality).

Certain effects of contextual evidence bias in Romero and Han () (their suggestion vs. contradiction uses) follow from the notion of ‘intent of a question’, an informal version of utility value. Furthermore, the authors treat Ladd’s readings in ()–() as genuine semantic ambiguity, as sketched in (). Under the outer reading, sentence () has the underlying syntactic representation (a), where negation is outside of the proposition double-checked by , and PPIs, not being in the immediate scope of negation, are licensed. Under the inner reading, () has the structure (b), where negation is inside the proposition double-checked by , and NPIs, being directly in the scope of negation, are grammatical. ()

Didn’t Amy arrive? a. LF for outer reading: [Q not [ [Amy arrived (already/too)]]] b. LF for inner reading: [Q  [not [Amy arrived (yet/either)]]]

Advantages of this analysis are that it distinguishes between original bias and evidence bias, that it correlates Ladd’s ambiguity with the distribution of PPIs/NPIs, and that it includes really-PosQs into the typology of biased question at no additional cost. It has two main problems. Problem (a) is that  is taken to contribute to the descriptive content of the question and, thus, it is in principle expected to be part of the descriptive content of the answer. But this is empirically incorrect: answering with a simple No to for example () under the outer negation reading conveys ¬(Jane is coming) rather than ¬(Jane is coming). Problem (b) is that the scopal combination ¬, crucial to explain outer HiNQs, is not attested when the two elements appear independently as not and really. Romero () and Gutzmann and Castroviejo Miró () suggest a potential way out of problem (a):  contributes not to the descriptive, at-issue content but to the

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

Conventional Implicature (CI) content of the utterance. This solves problem (a), since the CI content of a question is not part of what is being asked and, hence, it is not picked up by a normal straight answer. But it makes problem (b) more flagrant, given that CI content is not embeddable under negation (Potts ). Repp (, ) takes a different route. She distinguishes between Common Ground (CG) proper—the set of propositions that the speaker and addressee already agree upon— and an extended CG that functions as a provisional stack. V is defined as a CGmanaging operator, so that, when a sentence [Op p] is uttered, both the proposition p and the proposition  p are added to the provisional CG stack. The addressee can then react to those additions, tackling the addition of p and the addition of  p with different answer forms in German. This allows the author to circumvent problem (a). Furthermore, Repp and, following her, Romero () assume the (also CG-managing) operator  () instead of the scopal ordering ¬. This avoids problem (b) altogether. A remaining issue is how the underlying syntactic structures [Q  [¬p]] and [Q  p] predicted for inner and outer HiNQs respectively would combine with the notion of “intent of a question” to derive the desired empirical results. ()

λp.λws. 8w0 ∈ Epix(w) [ 8w00 ∈ Convx(w0 ) [ p 2 = Common-Groundw0 0 ] ]

Finally, we turn to the speech act line. In a first implementation, Krifka () concentrates on the evidence bias profile of PosQs, LowNQs, and outer-HiNQs. Speech act operators are defined as functions from input to output commitments (Merin ; Cohen and Krifka ). Two ingredients are key to the proposal. The first one is the speech act operator . While the question operator Q offers two possible continuations to the addressee and, thus, conveys no bias,  offers just one possible continuation to the addressee and is thus used in situations with evidence bias. The second key ingredient is speech act denegation, by which a speaker refuses to perform a certain speech act. The different evidence bias profiles of the three question types are then derived from the proposed underlying representations in ()–(). PosQs in () involve the operator Q and thus can be used in neutral contexts with no evidence bias. LowNQs in () are analyzed as requesting the assertion of the negative proposition ¬p and are, hence, incompatible with contextual evidence for p. Finally, outer HiNQs in () are analyzed as requesting a denegation, namely, as requesting that the addressee refuses to carry out the speech act of asserting p, which again makes this form incompatible with contexts involving contextual evidence for p. In a later version of this approach (Krifka ),  is substituted for a commitment operator ‘. ()

a. Is Jane coming? b. [Q [ Jane is coming]]

()

a. Is Jane not coming? b. [ [ [not [Jane is coming]]]]

()

a. Isn’t Jane coming (too/already)? b. [ [not  [Jane is coming]]]

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

, ,   



Advantages of this approach are that it is couched within a detailed framework of speech acts and discourse commitment space and that it derives the distribution of PPIs/NPIs. An important shortcoming is that it assigns the exact same semantic representation to LowNQs and to inner-HiNQs, thus predicting parallel behavior. But, as we saw in section .., the two forms have different bias profiles. Goodhue () builds on Krifka’s () proposal to derive mandatory original speaker bias for outer-HiNQs. Furthermore, he argues that original bias in PosQs like (b) with focus on the inflected verb arises from simple polarity focus in combination with Gricean reasoning. In a second implementation, Reese (, ) and Asher and Reese () use complex speech acts to derive original speaker bias and exploit nuclear pitch accents to establish rhetorical relations linked to contextual evidence bias, for example the relations of Confirmation and Counterevidence. To see just one case, outer HiNQs n’t p+PPI? are analyzed as complex speech acts of the form assertion•question, as follows. The final rising question intonation conveys that the speaker is not committed to p, that is, ¬committed(speaker,p); outer negation is metalinguistic negation carrying out a denial speech act, with the result that ¬committed(speaker,p) is denied and we obtain committed (speaker,p); finally, depending on whether the question is pronounced with a contentious nuclear accent like L*+H or not, we have the rhetorical relation of Counterevidence (e.g. in a context with evidence against p) or Acknowledgment (e.g. in neutral evidence context). The main advantages of this approach are that it derives both original and contextual evidence bias and that it correctly predicts the distribution of NPIs/PPIs. The main shortcomings lie on the compositional procedure. It is not clear, for example, how a metalinguistic negation embedded in an interrogative clause can scope over the contribution of the final rise.

... Cross-linguistic data High negation functions as a trigger of bias in questions in many languages. Romero and Han () provide data from German, Korean, and other languages. Other surface strategies triggering bias include: (i) particles like Spanish acaso (Escandell-Vidal ) and Japanese no and desho (Sudo ),⁷ (ii) word order in Italian dialects (Obenauer ), and (iii) prosody in English (Pierrehumbert and Hirschberg ), German (Pheby ; Kügler ) and Romance languages (e.g. Vanrell et al.  on Catalan).

.. T 

..................................................................................................................................

... Empirical characterization Tag questions are utterances consisting of a (declarative, imperative, or exclamative) clause, called the prejacent, supplemented with an (inflected or uninflected) short interrogative form or ‘tag’. In English, a variety of tag question forms is found, with the main types ⁷ Particles may also indicate lack of bias. For example, Hungarian -e indicates lack of contextual evidence bias (Gyuris ).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

classified as follows. First, in terms of phonological phrasing, the prejacent and the tag may form one single intonational phrase—post-nuclear tags in ()—or may be phrased separately—nuclear tags in ()–()—(Ladd ; Huddleston and Pullum ). Nuclear tags, in turn, vary in terms of polarity: the prejacent and the interrogative tag may have reverse polarities—reversed tags in ()—or the same polarity—constant tags in () (Cattell ; Huddleston and Pullum ; Malamud and Stephenson ). Finally, reversed nuclear tags vary in the contour of the interrogative tag itself, which may be rising, as in (a/bi), or falling, as in (a/bii) (Huddleston and Pullum ). This classification is schematized in (): ()

Main question tag forms in English post-nuclear tag: (30)

nuclear tag reversed tag: (31)

rising tag: (31a/bi)

() a. Paul is coming = isn’t he? b. Paul isn’t coming = is he? ()

a. Paul is coming, isn’t he? i. Paul is coming, isn’t he↑? ii. Paul is coming, isn’t he↓?

constant tag: (32)

falling tag: (31a/bii)

Post-nuclear tags

Reversed (nuclear) tags

b. Paul isn’t coming, is he? i. Paul isn’t coming, is he↑? ii. Paul isn’t coming, is he↓? () a. Paul is coming, is he? b. %Paul isn’t coming, isn’t he?⁸

Constant (nuclear) tags

In terms of pragmatic behavior, the empirical characterization of English tag questions in the literature has mostly concentrated on nuclear tag forms (but see Asher and Reese () and Northrup () on postnuclear tags) and is mainly based on two dimensions: (a) the attribution of the prejacent and (b) the degree of confidence or commitment expressed by the speaker on the prejacent.⁹ With respect to (a), the prejacent expresses the speaker’s belief or opinion in reversed tags, as in (), whereas the prejacent proposition is attributed to the addressee in constant

⁸ Many speakers reject negative constant tag questions like (b) (Huddleston and Pullum ). ⁹ Additionally, Northrup (: ) characterizes reversed tags in terms of original speaker bias and contextual evidence bias: Rising tags require original bias for the prejacent and falling tags contextual evidence for it.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

, ,   



tags, as in () (Cattell ; Huddleston and Pullum ; Malamud and Stephenson ; examples are slightly modified from the latter): ()

Context: A and B are discussing various traits of their mutual acquaintances. A thinks that Bill is attractive. B says, ‘I think Bill, more than anything else, is just a really nice guy’. A replies: A: He is attractive too, isn’t he? / #is he?

()

Context: A and B are gossiping. A doesn’t know anything about B’s neighbor. B says, blushing, ‘You’ve GOT to see this picture of my new neighbor!’. Without looking, A replies: A: He is attractive, #isn’t he / is he?

With respect to (b), reversed tags are taken to express different degrees of confidence via the prosodic contour on the interrogative tag, with a falling tune signaling a stronger commitment than a rising tune (Quirk et al. ; Huddleston and Pullum ; Asher and Reese ; Northrup ; Farkas and Roelofsen ). This is exemplified in ()–(), slightly abridged from Farkas and Roelofsen (). In () the speaker has some credence towards the prejacent but is in doubt and asks for verification; in (), the speaker is convinced of the truth of the prejacent and seeks the addressee’s agreement or acknowledgment: ()

Context: Amalia and Bert are a couple looking to furnish their new apartment. They make purchasing decisions together. They are looking at a table whose price is higher than what they would have expected. Amalia really likes it but she is not quite sure whether they can afford it. Amalia: It’s a bit too expensive for us, isn’t it↑?

()

Context: Amalia and Bert are a couple looking to furnish their new apartment. They are looking at a table that sells for $. They have agreed that they won’t spend more than $ on new furniture and they need other items besides a table. Amalia is convinced they cannot afford it. But Bert is sometimes inclined towards extravagance and keeps admiring this table. Amalia: It’s a bit too expensive for us, isn’t it↓?

Less consensus is found in the literature with respect to the degree of speaker confidence in constant tags. Huddleston and Pullum () maintain that they do not express doubt: the prejacent is typically repeated or inferred from something the addressee said.¹⁰ According to Northrup (: ff.), though, constant tag questions express the full gamut of degrees of confidence.

¹⁰ See also Quirk et al. (), who note that constant tags are often preceded by oh or so, as in (i), indicating that the speaker is making an inference based on what has been said in the conversation. I thank a reviewer for bringing up this point. (i) Oh / so he likes his job, does he?

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

... Theoretical modeling The pragmatic modeling of the phenomenon has mostly concentrated on nuclear tags. To capture (some or all) of their forms and pragmatic uses, three main lines of analysis have been pursued in the literature, two of which also applied to the analysis of biased polar questions in section : (I) the scoreboard discourse model, (II) the expressed proposition line, and (III) the speech act line. The first line uses the scoreboard discourse model (Lewis ; Farkas and Bruce ) to capture the behavior of all sub-types of nuclear tags. The scoreboard, an articulated view of the Stalnakerian common ground, distinguishes between commitments shared by the speaker and addressee—C(ommon) G(round) proper—, the discourse commitments of the speaker DCSp, the discourse commitments of the addressee DCAd, and the negotiation space called ‘Table’. To that, Malamud and Stephenson () add tentative or projected discourse commitments of speaker and addressee: DCSp* and DCAd*, respectively. Reversed tags are analyzed as adding the prejacent proposition p to the speaker’s DCSp* and constant tags as adding it to the addressee’s DCAd*, hence deriving the contrasts in ()–(). Furthermore, Farkas and Roelofsen () add the set of evidenced possibilities Evidencex of each discourse participant x, that is, the set of propositions which x has signaled to have some evidence for. This last layer is exploited to model the degree of confidence in p in falling vs. rising tags illustrated in ()–(): Falling tags add to EvidenceSp whereas rising tags add .¹¹ The expressed proposition line due to van Rooy and Šafářová () on biased polar questions is extended by Northrup () to account for the pragmatic function of falling vs. rising tags in ()–(). The analysis makes crucial use of the current and previous epistemic state of the speaker with respect to the proposition p in the prejacent. A falling tag signals that the speaker has an openness commitment to the propositions in [p or ¬p] at a previous epistemic state and (weak) commitment to p at the current state. This is taken to communicate that the speaker has developed a bias for p based on contextual evidence. In contrast, a rising tag signals (weak) commitment both towards openness and towards p at the current epistemic state, implicating that the speaker is open to either answer while putting more credence on p. Finally, we turn to the speech act line. In a first implementation, Asher and Reese () take reversed tag questions to carry out a complex speech act assertion•question. Falling intonation is associated with the rhetorical discourse relation Acknowledgment, hence the effect in (). Final rises block this discourse relation, leaving us with the rhetorical relation Confirmation, thus the use in (). In a second implementation within the speech act line, and following the framework laid out in Krifka (), Krifka () analyzes reversed and constant tags as carrying out coordinated speech acts. Reverse tags of the form [p, isn’t it?] are the disjunction of an assertion—committing the speaker to p—and a (monopolar) negative question—expecting that the addressee will commit to ¬p. Thus, the only discourse development that this form excludes is one where, at the same time, the speaker is committed to ¬p and the addressee ¹¹ For a comparison between question tag forms and rising declaratives within the scoreboard model, see, besides the works cited in the text, Jeong ().

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

, ,   



to p. This means that, if the addressee ends up committing to p, the speaker will as well. This, it is argued, corresponds to the use of reverse tags as (tentatively) putting forth a commitment to p and asking the addressee for support for it, as we saw in (). Constant tags are, instead, the conjunction of an assertion and a polar question: By using the constant tag [p, is it?], the speaker proposes to the addressee that both are committed to p. This, in turn, is argued to indicate that p is already understood as a commitment of the addressee, as we saw in ().

... Cross-linguistic data A look at other languages reveals further pragmatic functions of tag questions. Constant tags allow for different forms in Catalan (e.g. oi?, eh?) with slightly different discourse contributions (Castroviejo Miró ). Furthermore, certain tag forms in Cypriot Greek and German check not the truth of the prejacent but whether or not the illocutionary act intended by the prejacent has been accepted (Erotokritou ; Scheffler ).

.. R 

..................................................................................................................................

... Empirical characterization Rhetorical Question (RheQs) are interrogative utterances which, contrary to informationseeking questions (ISQs), have the feel of an assertion (Sadock , ; Progovac ; Caponigro and Sprouse ). Two examples are provided in (). (a) allows both for an ISQ reading and for a RheQ reading, with some phonetic and prosodic cues typically signaling the difference (Bartels ; Dehé and Braun ). (b), which contains a strong NPI like lift a finger or give a damn, only has the RheQ reading (Borkin ; Heim ). Useful tools to distinguish between RheQs and ISQs are the connective expressions after all and yet, which enforce the RheQ use (Sadock ), and phrases like I’m really curious or I really want to know, which enforce the ISQ use (Caponigro and Sprouse ), as illustrated in ()–(): ✓ISQ, ✓RheQ () a. Who helped you last night? b. Did John lift a finger to help you? * ISQ, ✓RheQ () a. Of course you should invite John! After all, who helped you last night? b. Of course you should not invite John! After all, did John lift a finger to help you? () a. I’m really curious: Who helped you last night? b. I’m really curious: # Did John lift a finger to help you? Two empirical points will be of importance. First, RheQs differ both from ISQs and from plain assertions in terms of subsequent discourse moves: While ISQs require proper answers like (A) (and do not accept reactions like (A0 )), RheQs allow for answers

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

but do not require them, as in (), and assertions forbid answers, as in () (Caponigro and Sprouse ):¹² ()

S: I’m really curious: Who helped you last night? A: Luca. A0 : # Ok, I shouldn’t complain.

()

S: After all, who helped you last night? A: Luca. / You are right, Luca. A0 : Ok, I shouldn’t complain.

()

S: After all, Luca helped you last night. A: # Luca. A0 : Ok, I shouldn’t complain.

Second, in terms of the content of the intended answer, RheQs with strong NPIs require that the (implicit or explicit) intended answer be negative, as shown in (). This contrasts with lexically unmarked RheQs like (a)/(S), which we saw allow for positive answers. ()

S: After all, who lifted a finger to help you? (Nobody. / #Luca).

... Theoretical modeling Two main research questions have guided the theoretical modeling of RheQs. The first one, concerning RheQs in general including lexically unmarked forms like (a)/(S), aims at deriving the assertive feeling (and the answer pattern) despite their interrogative form. The second research question concentrates on the role that strong NPIs like lift a finger play in enforcing the RheQ reading and a negative intended answer.

.... Assertive feeling of lexically unmarked RheQs Among the existing analyses of lexically unmarked RheQs in the literature, three main lines can be distinguished according to how the assertion-like interpretation is derived: (i) the semantic line, (ii) the speaker bias line, and (iii) the discourse line. In the semantic line, the assertive component of RheQs is part of the truth-conditional semantics (Sadock , ; Progovac ; Han and Siegel ; Han , a.o.). To see just one implementation, let us consider Han’s () analysis. Her ingredients to derive the rhetorical reading of the wh-question () are the following. First, a wh-word who{a,b,c}— where {a,b,c} is a silent restrictor—ranges over all combinations of elements of {a,b,c}, including the null element corresponding to the answer ‘nothing’, as sketched in (a).

¹² If an answer is provided to a RheQ, it may come from the addressee, as in (), or from the speaker herself, as in (i): (i)

S: After all, who helped you last night? Luca!

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

, ,   



Second, as a general heuristic for informativeness, if the speaker believes ¬P(x) to be likely of most/all x, then the most informative wh-question form will feature the opposite polarity and have the shape [whx P(x)] (cf. (ii)). Third, RheQs are assumed to come with certain auditory cues that signal that the interrogative should be semantically interpreted not as a question but as an assertion.¹³ Given this last ingredient, the interrogative sentence () cannot have the semantic denotation of a question—that is, it cannot denote the set of possible answers arising from all the possible values of who{a,b,c} in (a)—but has to have the semantic meaning of an assertion—that is, one single value must be selected for who{a,b,c} in the semantic derivation. The value selected will be the one that is consistent with the informativity heuristics above: Since the question form is [whx P(x)], the speaker must believe ¬P(x) for all x in the restrictor, and thus the value “nothing” is selected, as sketched in (b). ()

What has John done for you? [with rhetorical intonation] a. [CP What{a+b+c, a+b, b+c, a+c, a, b, c, nothing} [C’ has John done ti for you]] b. [CP What{nothing} [C’ has John done ti for you]]

An appealing characteristic of Han’s analysis is that it assigns a central role to auditory cues to derive the RheQ interpretation. However, since the analysis derives the same final truthconditions for a RheQ and for the corresponding assertion, it predicts the same subsequent discourse moves for both, contrary to ()–(). Furthermore, the assertion-equivalent interpretation obtained always has the opposite polarity to the question nucleus; yet () showed that positive intended answers to a positive RheQ are possible. We turn to the speaker bias line. Van Rooy and Šafárová () derive the rhetorical interpretation of questions like (a) as the limit case of speaker bias in polar questions (see section .). Semantically, the form p?, the form not-p? and the alternative question form p-or-not? all denote the set {p, ¬p}. Of the two propositions p and ¬p, the speaker chooses to express in the question nucleus of a polar question the proposition that has the higher expected utility, and she chooses the alternative question form if both propositions have the same utility, as defined in () above. The rhetorical use of (a), then, arises as an extreme case of bias via condition (ii): When the speaker has a firm original bias for ¬p (= ‘that you are not crazy’) but receives some contextual evidence for p (= ‘that you are crazy’), the speaker tries to resolve the conflict and chooses the question form p? because p being true would cause a wider revision than ¬p being true in her original belief-desire state. This approach correctly predicts that, in the bias conflict situation just described, (a) can felicitously be used as RheQ but (b,c) cannot. ()

[Situational context: The speaker originally believes that the addressee is not crazy, but the addressee is acting completely crazy right now. The speaker says:] a. Are you crazy? b. # Are you not crazy? c. # Are you crazy or not?

¹³ Han () assumes that RheQs have a falling final boundary tone, like assertions and unlike ISQ. But recent research locates the difference between RheQs and ISQs in edge tones and pitch accents (phonology) and in duration and voice quality (phonetics) (Dehé and Braun ).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

Given that (extreme) biased questions are semantically questions and not assertions, this analysis correctly expects different subsequent moves for RheQs and assertions. However, in terms of the content of the intended answer, this line wrongly predicts—just like the semantic line above—that the intended answer must invariably have the opposite polarity to that of the question nucleus. Furthermore, this analysis predicts that alternative questions can never be used as RheQs; this is contrary to fact, witness () from Biezma and Rawlins (): ()

Professor to the Ph.D. student complaining about the amount of work: Are you doing your Ph.D. or vacationing in Konstanz?

Finally, we turn to the discourse line. The aim here is to redefine the pragmatic import of interrogative utterances in discourse so that different effects—ISQ use, RheQ use, etc.—will arise depending on the particular discourse situation. According to Wilson and Sperber (), uttering an interrogative simply signals that the answer is relevant. Contextual factors will determine to whom: relevant to the speaker in information-seeking contexts and to the addressee in rhetorical contexts. In the same spirit, Caponigro and Sprouse () argue that, for any given interrogative Q, uttering Q leaves open whether or not the true answer to Q—that is, ⟦Q⟧w0—is already part of the addressee’s beliefs BelAd, part of the speaker’s beliefs BelSp, or part of the Common Ground CGSp+Ad shared by both. Depending on which of these possibilities is secured in the context, different pragmatic “flavors” will result, leading to the typology in ():¹⁴ ()

a. b. c. d.

Information-seeking: Quiz question: RheQ: Pondering question:

⟦Q⟧w0 ∈ BelAd, ⟦Q⟧w0 2 = CGSp+Ad ⟦Q⟧w0 ∈ BelSp, ⟦Q⟧w0 2 = CGSp+Ad ⟦Q⟧w0 ∈ CGSp+Ad ⟦Q⟧w0 2 = BelAd, ⟦Q⟧w0 2 = BelSp

Biezma and Rawlins () build on Caponigro and Sprouse () and, using the discourse scoreboard model from Farkas and Bruce (), distinguish two steps in updating the context with an interrogative utterance Q: (i) the speaker’s proposal to add ⟦Q⟧ to the Question-Under-Discussion (QUD) stack, and (ii) the actual addition of ⟦Q⟧ to the QUD stack. ISQs and RheQs are parallel with respect to step (i). They differ in that step (ii) leads to an inquisitive context in information-seeking uses—thus requiring an answer— and to a non-inquisitive context in rhetorical uses—thus not requiring an answer (though not forbidding one). Overall, the discourse line correctly derives different subsequent discourse moves for assertions, ISQs, and RheQs. Furthermore, the assertion-like behavior of RheQs is equally derived for intended answers of the same and opposite polarity to the question nucleus and for polar questions, wh-questions, and alternative questions. This makes this line very liberal. In fact, in its current form, this line is too liberal: As we saw in (), in certain ¹⁴ (i) exemplifies quiz questions and (ii) pondering questions: (i) [Teacher to student:] Now your turn, Lucía. What’s the capital of Baden-Württemberg? (ii) What is the meaning of life?

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

, ,   



discourse situations with intended answer ¬p, the opposite polarity form p? must be used and the forms not-p? and p-or-not? are disallowed. It needs to be determined how this line can be constrained to delimit the potential pairings of question form plus intended answer.

.... The effect of strong NPIs Recall from (b)–(b) and () that interrogatives like () with strong NPIs mandatorily lead to a RheQ reading and only allow for negative (intended) answers. ()

Did John lift a finger to help you?

(=b)

Two main accounts currently compete to derive these facts.¹⁵ Both of them assume that strong NPIs are accompanied by an (often silent) even, defined in (), but each exploits a different presupposition of even: ()

a. [evenC IP] presupposes that: (i) for every q in C other than ⟦IP⟧, q is more likely than ⟦IP⟧. (ii) for every q in C other than ⟦IP⟧, q is true. b. [evenC IP] asserts ⟦IP⟧.

The first analysis, due to Guerzoni (), capitalizes on the presupposition (ai): the proposition ⟦IP⟧ that even combines with is presupposed to be unlikely or “hard” (compared to other alternative propositions in a salient set C). To try to satisfy this presupposition in example (), where the IP contains the bottom-of-the-scale term lift a finger, even must scope above the trace of whether, which ranges over the polarity functions ‘yes’ (λp.p) and ‘no’ (λp.¬p). This leads to the LF (). The resulting interpretation is the set of propositions in (), with the presuppositions underlined. Crucially, the presupposition of the negative proposition is satisfied given world knowledge, but the presupposition of the positive proposition is not. This means that there is only one possible felicitous answer for the addressee to choose—hence, the mandatory RheQ reading—and that this unique felicitous answer is negative—hence, the mandatorily negative (intended) answer. ()

LF of (): [CP whetheri Q [IP even [IP ti [John lifted a finger to help you]]]]

() { λw: John doing a tiny amount of work to help you is hard. John did a tiny amount of work to help you in w, λw: John not doing a tiny amount of work to help you is hard. John did not do a tiny amount of work to help you in w } The second analysis, due to van Rooy (), capitalizes on the presupposition (aii) of even with declaratives and extends it to even with interrogatives. Just as the declarative (a) presupposes that all other alternative propositions in a salient set C are true, the ¹⁵ See previous accounts by Ladusaw (b) and Gutiérrez-Rexach () containing some seminal ideas.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

interrogative (b) presupposes that all other alternative questions in C are settled, that is, resolved in the CG. In the case of our example (), this means that, for any amount of help x other than lifting a finger, the question Did John do x to help you? is already settled in the CG, and that the only unsettled question is whether John did the minimal amount of work (=lifting a finger) to help you or not even that. Since (just) lifting a finger amounts to not doing anything substantial at all, sentence () communicates that John did not do anything to help you—hence mandatorily deriving the rhetorical effect and the negative (intended) answer. ()

a. Mary can evenC speak FRENCH. b. Can Mary evenC speak FRENCH?

.. S

.................................................................................................................................. We have seen that, besides raising an issue (e.g. {p, ¬p}), a number of interrogative clause types express a bias or (tentative) commitment on the speaker’s side. First, different polar question forms, while (possibly) expressing the same truth-conditional content, carry different pragmatic bias profiles resulting from a combination of original speaker bias and contextual evidence bias. Second, tag questions vary in the attribution of the proposition expressed by the prejacent and in the degree of endorsement of the speaker towards that proposition. Finally, polar, wh- and alternative questions can be used when the issue raised is already resolved in the Common Ground, leading to a rhetorical interpretation; when the question form includes a strong NPI, the licensing conditions of this NPI—via even—enforce the rhetorical reading and a negative intended answer. To capture the pragmatic nuances of all these question types, a more articulated view of the discourse is needed. Current approaches put the burden of the explanation on the status of the relevant proposition in the epistemic state of the speaker, on Common-Groundmanaging operators like , on complex speech acts, or on refined versions of Lewis’ () scoreboard model of discourse.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  ......................................................................................................................

  ......................................................................................................................

 

.. I

.................................................................................................................................. I natural language, standard sentential negation is a truth-value reversal operator whose semantics can be generalized to a complement-set operation. Standard negation has a rich syntactic and pragmatic dimension (Zanuttini ; Horn , ). At the interface between syntax and the systems of interpretation, a phenomenon that deserves special attention is the so-called ‘expletive’ negation (EN), corresponding to cases in which a negative formative that is often (though not always) linguistically indistinguishable from standard sentential negation is used in main and embedded clauses without providing, according to the received wisdom, any truth-conditional contribution to interpretation. For the reader’s convenience, some uses of EN are exemplified below for languages as different as Italian (embedded comparative clause; Napoli and Nespor ), German (temporal clause introduced by bevor; Krifka ) and Korean (negative selected complement clause of the verbal predicate hope; Yoon ):¹ Carlo () Maria è più alta di quanto non sia Maria is taller than not is- Carlo ‘Maria is taller than Carlo’ umzog () Mozart vollendete nicht sein Requiem bevor er nicht nach Wien Mozart finished not his Requiem before he not to Vienna moved ‘Mozart did not finish his Requiem before he moved to Vienna’ kitayha-koissta Mary-ka anh-oci-anh-ul-kka () John-un John- Mary- -come--- hope- ‘John hopes that Mary might not come (although it is unlikely to happen)’

¹ According to Yoon  (p. ), “the negative interpretation comes from the first anh (Neg: real negation) while the second anh (Neg: EN) is logically vacuous.”

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

EN has a number of puzzling properties: (i) it has a variable cross-linguistic distribution; (ii) it is optional intralinguistically (its use is generally not compulsory in the structures in which it is admitted); (iii) it does not reverse sentence polarity; (iv) contrary to antiadditivity operators such as standard negation it does not license items of negative polarity (NPIs) or negative concord words (NCIs) with an indefinite interpretation (i.e. it is incompatible with negative-concord (NC)). Moreover, when it co-occurs with standard negation (cf. () above), it does not give rise to the standard semantics of double-negation (DN). According to the available typological data, EN appears to be preferably licensed in non-veridical contexts in the subjunctive mood and is most typically found in the following syntactic environments: (i) in the complement of verbs expressing fear, prohibition, hindering, avoidance, denial, doubt and, though more restrictively, hope (as in Korean/ Japanese); (ii) in clauses introduced by specific complementizers such as until, without, unless, etc.; (iii) in temporal clauses introduced by before (but not by after); (iv) in comparative and exclamative clauses. On these grounds, the question arises whether EN is a vacuous morpheme (a really ‘expletive’ element), whether it somehow loses its negative import as an effect of syntactic derivation and compositional interpretation, or whether there are ways to interpret it as a real negation, after all. In this chapter, I will mainly concentrate on these questions, that is on the puzzles raised by EN’s nature and interpretation. The chapter is structured as follows. In section .., I will discuss some of the most common analyses of EN, and more particularly: (i) the syntactic relationship between EN and NC; (ii) the proposals according to which EN is a real negation; (iii) the idea that EN is a special formative linked to an additional evaluative/expressive layer in the semantics of language. In section .., I will inquire into the peculiar behavior of EN in Italian temporal, exclamative, and comparative clauses, as a (possibly revealing) individual case study. In section .., I will summarize the results of the preceding discussions while pointing to some new promising directions of inquiry.

.. W  EN? T ---

.................................................................................................................................. In the literature, EN has been argued to be similar to NC in a number of respects. For instance, both phenomena have been claimed to involve expletiveness (EN-contexts are interpreted positively, while in NC-contexts there is one or more morphologically negative elements that are interpreted positively) and optionality (EN is systematically optional, while NC is at least partly optional, in that, for instance, an NCI in subject position may or may not co-occur with sentential negation giving rise to NC). Moreover, both EN and NC involve clause-boundedness and, arguably, a local dependency from a head hosting a negative feature of some sort (expressing either a non-veridical operator (EN) or an anti-additive operator (NC); Espinal , , ). However, there is also clearly a point at which the parallelism stops. First, NC involves a dependency from anti-veridical operators, whereas EN most typically involves a

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 



dependency from non-veridical operators,² and can arguably surface even in veridical contexts, as in the case of until-clauses in Spanish and other languages and in the case of time measure constructions with since in languages such as Korean (Cepeda ).³ Second, NC has been related to scope-marking effects, whereas EN has been usually taken to be scopally inert and not to exhibit, more generally, any significant interaction with NCIs and NPIs (see however Delfitto, Melloni, and Vender  for the claim that EN is in fact scopally active in Italian). Third, and perhaps more noticeably, NC is definitely not semantically vacuous as a grammatical phenomenon (Déprez et al. ), whereas it is traditionally claimed that EN is, though this chapter will present important evidence to the contrary. Moreover, as I have already observed, the set of licensers is sharply distinguished for NC and EN: anti-veridical operators for NC and non-veridical (or even veridical) contexts for EN. Given the set of properties generally assigned to EN, it should be emphasized that there is abundant room for cross-linguistic variation. For instance, both the clausal complement of ‘doubt’ and before-clauses qualify as non-veridical operators. However, whereas abans ‘before’ and dubtar ‘doubt’ license the expletive marker no in Catalan, this is not the case for a closely related language such as Spanish (Espinal ). For instance, a sentence such as () in Catalan cannot be interpreted in Spanish (or in Italian) as indicated in the English translation; in these languages, the negative marker is rather interpreted as a real sentential negation: () Dubto que no mengi I doubt that not eat- ‘I doubt that he eats’ The idea that verbs like doubt and prepositions like before contain a syntactically active Neg-feature interacting with the Neg-feature on EN has been extensively defended in a series of contributions by M. T. Espinal (Espinal , , ). Roughly, the insight is that heads qualifying as non-veridical operators are syntactically ‘negative’ and that their Neg-feature attracts (i.e. checks and absorbs) the Neg-feature expressed by EN. The existence of this type of syntactic dependency is corroborated by the observation that Catalan abans or Spanish antes ‘before’ easily license, besides the expletive negative marker no, also NCIs as indefinites, as shown in () below: ()

a. Abans que ningú digui res, deixeu-me donar-vos la benvinguda

(Catalan)

b. Antes de que nadie diga nada, dejen que les dé la bienvenida ‘Before anyone says anything, let me welcome you’

(Spanish)

Syntactic accounts in terms of Neg-feature attraction raise a number of delicate issues concerning the distribution and the interpretation of negative features and, more generally,

² Simplifying a bit, we can state that a propositional operator O is veridical if and only if Op entails or presupposes that the proposition p is true. Conversely, an operator O is anti-veridical iff Op entails that not p is the case. If O is neither veridical nor anti-veridical, it is said to be non-veridical (cf. for instance Giannakidou b). ³ I am indebted to an anonymous reviewer for drawing my attention to these data.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

the interface between morpho-syntax and the systems of interpretation, both conceptually and empirically. Conceptually, non-veridical contexts as before-clauses and verbs of doubt are not negative contexts; rather, they qualify as neutral contexts in which there is no commitment to the truth of the relevant proposition on the part of the speaker. Why should then the Neg-feature expressed by the relevant heads in EN-contexts be treated on a par with the Neg-feature expressed in NC-contexts, which is promptly interpreted as expressing anti-additivity (i.e. negation proper)?⁴ Empirically, the interesting question arises whether the negation licensed in before-clauses is an instance of EN or reflects the presence of a real negation in the underlying logical form. As a matter of fact, syntactic structures of the form ‘A before B’ can be analyzed as involving universal quantification over times and a negative interpretation of the predicate expressed by the before-clause (as in Krifka , partially based on Anscombe ): “For every time t preceding A, ¬B(t).” Informally, the before-clause is negated at all times preceding the time at which the main clause is evaluated; this is equivalent to a non-factual (i.e. non-veridical) analysis of beforeclauses (Krifka  a.o.). A convenient exemplification is given in (), where (a) is assigned the logical form in (b), roughly interpreted as indicated in (c): ()

a. Warn her before she causes any trouble! (warn her ¼ A; before she causes . . . ¼ B) b. A(τ) ∧ ∀t: t < τ. ¬B(t) ≅ A(τ) ∧ ¬9t: t < τ. B(t)⁵ c. Ensure that all times t preceding the moment τ at which you warn her are such that she does not cause any trouble at t

On these grounds, one might adopt the view that EN in before-clauses, as in the Italian equivalent of (a), given in () for the reader’s convenience, is not really expletive but somehow reflects the negative interpretation of the before-clause in logical form: ()

Avvertila prima che non combini qualche guaio!

The question is: can this non-expletive analysis be generalized beyond before-clauses? There is in fact a tradition of thought, in linguistics, according to which EN is more strictly tied to real negation than the expletive account would be ready to admit. According to this tradition, the EN occurring in an embedded clause selected by a verb or noun expressing fear, prohibition, hindering, avoidance, denial, or doubt expresses the negative content of the superordinate predicate. The claim is in fact that all these are predicates with a negative import. Similar ideas have been put forward with respect to a large variety of languages, such as Old/Middle English (Jespersen ), French (Muller ), Polish (Jablonska ), Russian (Brown and Franks ) and in general, the approach is best known as involving the concept of ‘paratactic’ negation, introduced by Jespersen. It is probably within this tradition of thought that Seuren  proposed that comparative ⁴ It is generally assumed that strong polarity items cannot be licensed by downward-entailing operators, but only by anti-additive operators, that is, sentential negation and negated existentials such as no-N and never. Formally, an operator O qualifies as anti-additive if and only if O(A ∨ B) is logically equivalent to O(A) ∧ O(B). ⁵ For simplicity, the formula refers to τ as the referentially given time of the event expressed by the main clause.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 



clauses also actually involve semantic negation: a sentence like John is taller than Mary should be read as John is tall at a degree d, and Mary is not tall at that degree. No surprise, then, that in languages such as Italian one can say things of the sort John is taller than Mary is not, as exemplified in () above. Again, the use of EN would be related to the presence of a real negation in the underlying logical form. Unfortunately, the evidence available by now does not support the claim that in all contexts featuring EN there is a real negation semantically, either in the main or in the embedded clause. In effect, nowadays the perhaps most accepted style of analysis of EN relates its use to the activation of a distinct expressive-evaluative layer of semantic interpretation (Potts ). From this perspective, EN is certainly not an instance of real negation. On the contrary, EN and real (i.e. logical) negation are two homophonous elements, each relevant for a distinct layer of meaning. The evaluative interpretation of EN requires the use of doxastic models reflecting the epistemic stand of the speaker (whether, say, the relevant proposition is considered (un)likely or (un)desirable by the speaker) and represents the reflex of grammaticalization of perspective and subjective mood, further exemplified by predicates of personal taste (Lasersohn ), mood choice (Quer ; Giannakidou ) and other phenomena of the same kind. Technically, the evaluative sense of EN should be understood as an utterance modifier, in the form of a conventionalized implicature (CI; Yoon ). It should be emphasized that this analysis of EN is more related to the tradition of thought that takes EN to be a truthconditionally vacuous element than to the analysis, mentioned above, involving the notion of ‘paratactic’ negation. The reason is that the truth-conditional content of what is said and the expressive/evaluative content are strictly separated from each other. According to Potts’s approach (multidimensionality of CIs), the meaning contribution of EN is dealt with as a separate dimension with respect to the basic semantics of the sentence containing EN. In other words, EN contributes to the ‘enriched’ meaning of the sentence, but its contribution is logically and compositionally independent of what is said (as confirmed by the observation that CIs are scopeless). As an exemplification of this expressive/evaluative interpretation of EN, consider first the Korean sentence in () above: whereas the non-factive complementizer simply expresses uncertainty with respect to the propositional content of what is hoped, the insight is that adding EN turns uncertainty into unlikelihood.⁶ In other words, the CI triggered by EN expresses the speaker’s epistemic stand according to which the speaker hopes that p though she knows that it is extremely unlikely that it will be the case that p. The semantic value of the relevant CI further translates, in different contexts, in a large variety of different emotional contents, often achieved by emphasizing the meaning expressed independently of EN. According to this line of analysis, exclamative sentences such as those in (a) and (b), containing EN, further emphasize the sense of surprise, independently expressed by the semantics of the wh-exclamative, for the fact that the agent is really doing everything. Similarly, before-clauses of the sort indicated in (c) are likely to express, when containing EN, the undesirability of the propositional content ⁶ Yoon () strongly emphasizes (p. ) that “both Japanese and Korean clauses with EN must take a non-factive complementizer (NFcomp)” and that (p. ) “it is certainly not coincidental that the nonfactive complementizer ci/kka in Korean and ka in Japanese are in an identical form to a question particle.”

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

expressed by the before-clause (Yoon ). More generally, these examples show that the CIs tied to the use of EN directly trigger the activation of hierarchies of likelihood/ desirability and indirectly trigger the expression of the expressive/emotional values conveyed by the reference to these hierarchies:⁷ () a. Was du nicht alles machst! what you -all do ‘The things you do!’ b. Che cosa non ha fatto Gianni! what thing not has done Gianni ‘The things Gianni did!’

(German: Roguska )

(Italian: Delfitto and Fiorin )

c. Before he burns down any house . . . , before he kills anyone . . . , etc. Interestingly, the original analysis of EN in Italian comparatives (Napoli and Nespor ) was also based on the insight that EN expresses unlikelihood in these contexts. For instance, in the case of () above, what the use of EN would convey is the sense of unlikelihood of the situation according to which Carlo’s height comes close to Maria’s. Clearly, all these predictions should undergo some more detailed empirical scrutiny. This is what I intend to do in section ., by assessing the predictions made by the evaluative approach for temporal, comparative, and exclamative clauses in Italian. This will lead us to the formulation of an interesting alternative analysis.

.. EN  , ,   

..................................................................................................................................

... Temporal clauses The evaluative analysis of EN takes EN as expressing an evaluative relation between an individual x and a proposition p. The evaluative relation generally consists of an attitude that x has towards p, corresponding to the epistemic state x is in. If we apply this analysis to before-clauses, we get as a result that the propositional content expressed by the before-clause is evaluated as unlikely or as undesirable, in terms of the conventionalized implicature triggered by EN. Undesirability is arguably the expressive meaning readily associated with

⁷ An anonymous reviewer observes that the “undesirability interpretation seems to be related more to world knowledge than to the presence of the negative marker.” However, there need not be any contrast between the presence of EN and access to knowledge of the world: in most cases, the whole point is that there is a syntactic marker of undesirability, in full agreement with the contextual interpretation of the sentence. Whether undesirability covers all the uses of EN in languages such as French or Italian is a separate issue, and in fact I should strongly emphasize that the present chapter does not defend the expressive/evaluative account as the correct treatment of EN at a general cross-linguistic level. In fact, I will argue for a possible reduction—in many if not all languages—of the desirability/likelihood interpretation to the mechanism of implicature cancellation that will be discussed in section ..

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 



the before-clauses in (c), as well as the meaning arising in counterfactual uses of a before-clause, exemplified in () (Yoon ): ()

Mary defused the bomb before it exploded

Surely enough, not all before-clauses are interpreted counterfactually. In fact, it is generally accepted that what ‘A before B’ entails about B is largely determined by the context of utterance. In order to see this in some detail, let us consider for instance a sentence like (): ()

I arrived (at the party) before you arrived

By generalizing the truth-conditions exemplified in (b), the semantics associated with () simply holds that at all times t preceding my arrival, you had not arrived at the party yet.⁸ However, uttering () is likely to trigger the conversational implicature according to which you arrived at the party after me. Notice that whenever uttering a before-clause activates this layer of implicated meaning, the propositional content of the before-clause is interpreted factually. In (), what is meant is that you actually arrived, though this happened only after my arrival. Of course, this conversational implicature (together with the concomitant factual interpretation) can be cancelled. This is obviously the case in counterfactual contexts like (): if Mary defused the bomb, the bomb never exploded. Consider also the instance of before-clause in (a): ()

a. Bring the dog to the vet before he dies!

The logic of the order/advice expressed by (a) is most plausibly that bringing the dog to the vet will be crucial to avoid that the dog dies, not only before but obviously also (in a pragmatically relevant period) after your visit to the vet. There is consequently no implicated meaning to the effect that the dog dies after you brought him to the vet. As a result, the before-clause in () is not interpreted factually. What this suggests is that if a before-clause is interpreted factually depends on whether the cognitive/pragmatic conditions of utterance legitimate the conversational implicature that may be associated with the before-clause, that is the proposition according to which, given ‘A before B’, B actually took place after A. On these grounds, it is quite tempting to suggest that EN is a way of syntactically encoding the process of implicature cancellation. Consider for instance the Italian counterpart of (), which quite naturally supports the use of EN, as shown in (b): ()

b. Porta il cane dal veterinario prima che non muoia

Arguably, what EN expresses in (b) is that the conversational implicature that might be entertained, according to which the dog will die after your visit to the vet, should not be entertained at all. Technically, this means that EN applies to the implicated proposition The dog will die after you’ve brought him to the vet, acting as a truth-reversal operator: It is not ⁸ Cf. Krifka (), which also presents an original and intriguing analysis of the relation between the asserted and implicated meanings conveyed by before-clauses.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

the case that the dog will die after you’ve brought him to the vet. The conceptual advantage of this analysis is that EN preserves the semantics of a real negation, though it shifts its domain of application from the proposition expressing the asserted meaning to the proposition expressing the implicated meaning (cf. Moeschler  for similar ideas concerning metalinguistic negation). Interestingly, there is a significant intersection with the evaluative analysis of EN. The fact that EN in (b) encodes implicature denial is perfectly aligned with the undesirability of the state of affairs expressed by the proposition The dog will die after your visit to the vet. However, the two analyses are also very different. In the evaluative analysis, EN syntactically (pre-)encodes a specific conventionalized implicature, with respect to a hierarchy of likelihood/desirability. In the implicature-denial analysis, EN applies—as a truth-reversal operator—to the conversational implicature that is independently generated by the beforeclause that hosts EN. As a consequence, in the evaluative analysis EN does not retain the semantics of a real negation, whereas this is clearly the case in the implicature-denial analysis, in which EN is a truth-reversal operator whose domain of application is the implicated proposition. Given the conceptual attractiveness of the implicature-denial analysis, the question is now whether there is substantial empirical evidence that might independently support it. Consider in this regard the minimal pair in (), modeled after a similar contrast discussed in Del Prete (): ()

mia madre a. Mio padre parlerà prima che non lo faccia my father will talk before that not it does- my mother ‘My father will talk before my mother does it’ b. Mio padre è nato prima che (*non) nascesse mia madre my father is born before that not was-born- my mother ‘My father was born before my mother was born’

The use of EN in (b) gives rise to utter unacceptability in Italian. Significantly, whereas it makes sense to negate the implicated proposition associated with the before-clause in (a) (i.e. My mother will talk after my father), it makes no sense at all to negate the implicated proposition triggered by the before-clause in (b) (i.e. My mother was born after my father), since this would give rise to the reading according to which my mother was born neither before nor after my father, that is her birth never took place, a statement conflicting with the standard encyclopedic knowledge that is part of the conditions of utterance of (b). In plain words, (b) requires a factual interpretation of the before-clause, and this requirement is incompatible with the non-factual interpretation triggered by EN through implicature-denial. This correctly predicts that the use of EN is not legitimate in (b). As a point of further empirical corroboration, consider now the before-clauses in (), discussed in Delfitto, Melloni, and Vender (): ()

a. Se continua così, morirà prima di (*non) fare testamento ‘If he goes on like that, he will die before he makes a will’ b. Fai testamento, prima di (*non) morire! ‘Make a will, before you die!’

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 



These are also contexts in which EN is sharply ruled out in the before-clause.⁹ In (a), the asserted meaning is that he will not make a will at any time t preceding his death. Clearly, there is no implicated meaning to the effect that he will make a will after his death (under the view that conversational implicatures are costly cognitive processes that are triggered only in cognitively supportive contexts).¹⁰ There is thus nothing for EN to apply to, and EN is ruled out as deviant. In (b), the point about the order/advice expressed is that the interlocutor should avoid dying at any point t preceding the moment at which he makes a will. Certainly, he will die at some point after making a will. Therefore, negating the implicated proposition according to which he will die after making a will results in awkwardness. In this case, the advice expressed would be to avoid dying both before and after making a will, and this is clearly not what the advice is meant to express. These observations can be further elucidated by comparing (b), where EN is ruled out, with (b), where EN is fully legitimate in the before-clause. The whole point reduces to the remark that EN negates the implicated proposition triggered by the before-clause, a reasonable move in (b) (we intend, and in fact wish, that the dog does not die for a long time after the visit to the vet) but an utterly unreasonable move in (b) (the addressee is not urged to avoid dying after making a will, he is simply urged to make a will before his death).

... Exclamative clauses Let us consider now EN in exclamative clauses, on which there is a rich literature (cf. Eilam  for Hebrew, Meibauer  and Roguska  for German, Zanuttini and Portner  for Paduan). Here, I will concentrate on the evaluative approach, according to which realizing EN in exclamatives is tantamount to strenghtening the surprise effect independently expressed by the exclamative. The Italian wh-exclamative in (b), for instance, is supposed to strengthen the surprise effect triggered by the things Gianni did by emphasizing the unlikelihood of the relevant propositional content, as a function of the semantics of EN. The phenomenon is also attested in other kinds of exclamative sentences, as shown by () in English, featuring two negation markers (Yoon ): ()

Aw, you didn’t not call the plumber!

Suppose () is uttered in a context in which someone expected her interlocutor to call the plumber to fix a broken sink. According to the evaluative approach, the surprise effect is

⁹ An anonymous reviewer suggests that the (non-)acceptability of EN might simply be sensitive to the alternation between infinitival and subjunctive forms. Arguably, this is not the case. For instance, EN in (b) remains acceptable even in the infinitival variant of (b): ‘Se fossi in te, proverei a vedere un medico prima di non morire’ (If I were you, I’d try to see a doctor before you don’t die). Conversely, EN in (a) remains awkward even in the subjunctive variant of (a): ‘Se continui così, morirai prima che tu (??non) riesca a fare testamento’ (If you go on like this, you’ll die before you don’t succeed to make a will). ¹⁰ Here and in the rest of the present chapter, I will assume a post-Gricean model of conversational (scalar) implicatures, along the lines of Relevance Theory (Sperber and Wilson ).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

compounded in () by the presence of EN, which further underlines the unlikelihood of the relevant negative proposition, that is, You didn’t call the plumber. Let’s go back to the wh-exclamative in (b). According to the standard approach to the semantics of wh-exclamatives, based on the semantics of questions, the relevant exceptional sentence (the particular exceptional thing that Gianni did) must be selected within a set of algebraically structured propositional candidates, which are hierarchically ordered in terms of informativity (cf. Delfitto and Fiorin , and the references cited therein). To exemplify, suppose that in the context of (b) there are basically three propositions expressing what Gianni may have done: p, q, z. This leads to the propositional hierarchy in (): ()

Gianni did p, q, z Gianni did p, q; Gianni did p, z; Gianni did q, z. Gianni did p (or q, or z)

In uttering the variant of (b) without EN (‘Che cosa ha fatto Gianni!’), the speaker is asserting that there are one or more things that Gianni did, corresponding to exceptional/ surprising propositions. The speaker is also presumably implicating that the stronger options expressed by the hierarchy in () are not selected (these correspond to the more informative, hence less likely, propositional variants). In other words, though asserting the variant of (b) without EN is truth-conditionally compatible with the fact that Gianni did p, q, z (i.e. everything), it is implicated that this is not the case: as in the traditional analysis of scalar implicatures, the stronger options are denied. Suppose now further that the role of EN in (b) consists in negating the conversational implicature triggered by a whexclamative, as was the case, by hypothesis, with before-clauses. In the case of (b), the implicature consists in the denial of the stronger options within the propositional hierarchy in () (Gianni did p, q . . . Gianni did p, q, z), to the effect that negating the implicature is tantamount to lifting this denial: in uttering (b), the speaker is asserting that Gianni did one and possibly all the things expressed by the propositions in () (all of them corresponding to exceptional/surprising propositions). In plain words, EN lifts the negative implicature triggered by the variant without EN (he did some but not all the things he might have done), and as a result (b) gets the universal flavor according to which the speaker is expressing surprise for the fact that Gianni virtually did all the amazing things one might conceive of. In this way, the strengthened surprise effect yielded by EN in (b) is simply a by-effect of the core semantics of EN (i.e. implicature denial).

... Comparative clauses Can this analysis be extended to comparative clauses? Interestingly, the first influential analysis of EN in Italian comparatives (Napoli and Nespor ; cf. also Donati ) is quite close to the spirit (if not to the letter) of the evaluative approach, since it takes a sentence like (), reproduced here as () for the reader’s convenience, to involve the presupposition, on the part of the speaker, that it is unlikely that Carlo’s height comes close to Mary’s: ()

Maria è più alta di quanto non sia Carlo

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 



If we assume that this insight is empirically well-motivated, the question that arises is whether this interpretive effect can be derived from the core semantics of EN as an implicature denial operator, extending the analysis proposed above to the case of comparatives.¹¹ At first sight, this seems problematic, since implicature cancellation is somehow inherent to the semantics of comparatives, to the effect that there should be nothing for EN to apply to. This is shown by the contrast between the utter unacceptability of (a) and the perfect status of (b): ()

a. #Mary is tall, though she is in fact short b. Mary is taller than Carlo, though both of them are in fact short

The asserted meaning of Mary is tall is the hardly informative proposition that there is a degree d such that Mary is d-tall. Informativity is thus plausibly achieved by adding the implicated meaning according to which d is higher than the (contextually determined) average. When asserting that she is tall, the speaker implicates in fact that she is taller than the average, and that’s the reason why (a) sounds contradictory.¹² Conversely, the comparative structure in (b) does not need any implicated meaning to achieve informativity: what matters is the comparative judgment according to which the degree d such that Mary is d-tall is higher than the degree d’ such that Carlo is d0 -tall, independently of the further piece of information concerning the relation of the degrees d and d0 with the average. This explains why (b) is perfectly acceptable: the information according to which d is higher than d0 stands even in contexts where both d and d0 are low in the scale. If implicature cancellation is part and parcel of the semantics of comparative clauses, there is no implicated proposition. So, what does EN apply to? We might propose, as is usually the case within the evaluative approach to EN, that EN simply has a strengthening effect with respect to an independently present reading: in this case, it would syntactically (pre-)encode implicature cancellation. However, strengthening an independent process of implicature cancellation is not the same as performing implicature denial, and EN in temporal and exclamative clauses arguably does the latter, not the former. This last observation suggests in fact a more explanatory and principled analysis. In a nutshell, the basic insight is that the difference between the variant of () without EN and () is that the variant without EN simply involves implicature cancellation, at a global level; whereas EN in () encodes the denial of the implicated proposition, at a local level.¹³ ¹¹ See Delfitto, Melloni, and Vender () for a different empirical generalization. ¹² Clearly, this is tantamount to deriving the ‘evaluative’ interpretation of tall in terms of a conversational implicature. ¹³ For a localist view of scalar implicatures see e.g. Chierchia, Fox, and Spector (). The difference between calculating a scalar implicature ‘globally’ or ‘locally’ can be conveniently illustrated by considering the example in (i), where the scalar term triggering implicature calculation is in italics: (i) Every student passed some of the exams When the implicature is calculated ‘globally’, the result obtained is (ii): (ii) It is not the case that every student passed all the exams When the implicature is calculated ‘locally’, the result obtained is (iii): (iii) Every student passed some but not all the exams Clearly, (ii) and (ii) have different truth-conditions.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

Implicature denial at the local level is something different than implicature denial at a global level. In other words, the meaning of EN in () is that of performing a special kind of implicature denial (locally) rather than that of strengthening the cancellation of an already existing implicated meaning (globally). The relevant difference is shown in (): ()

a. The degree d such that Maria is d-tall is higher than the degree d0 such that Carlo is d0 -tall, and it is not the case that d and d0 are higher than the average. b. The degree d such that Maria is d-tall is higher than the degree d0 such that Carlo is d0 -tall, and it is not the case that d0 is higher than the average.

This means that EN, as realized in embedded comparative clauses, brings about a process of local implicature denial, according to which what is negated is the implicature according to which the degree associated with the embedded clause is higher than the average. This analysis has many advantages. First, it explains the presuppositional flavor we were interested in (Carlo’s height is unlikely to come close to Maria’s height), through the proposed ‘local’ process of implicature denial: since the implicated proposition that Carlo is taller than the average is (locally) negated, the odds are clearly not in favor of Carlo being taller than Maria. Second, EN has not simply a strengthening effect with respect to the global process of implicature cancellation, since in fact EN applies locally (i.e. it applies to a distinct propositional content). This is compatible with the proposal concerning the core semantics of EN that has been defended in the present section, according to which EN is a polarity-reversal operator that applies at the level of implicated meaning. Third, and perhaps more noticeably, the account proposed for EN in comparatives opens the possibility that local implicature denial applies in entirely different syntactic contexts featuring EN. A case in point might be the complements of verbs of fear, exemplified by the French structure in (), where fear is expressed for the eventuality that he comes: ()

Je crains qu’il ne vienne I am afraid that he not comes ‘I’m afraid that he comes’

According to the evaluative approach, the use of EN in () triggers the canonical unlikelihood/undesirability scale. Here, what is expressed is that I am afraid for the eventuality that he comes, a highly undesired eventuality. What about the analysis of EN in terms of implicature denial? A reasonable implicature associated with verbs of fear is that if someone is afraid of something, it is because there is a certain degree of probability, or even of likelihood, that that something takes place. The implicated proposition is thus (a), whereas its denial corresponds to (b): ()

a. It is likely that he comes b. It is unlikely that he comes

Interestingly, the operation of global implicature denial shown in (b) exactly corresponds to the interpretive effects linked to the use of EN with verbs of fear in Korean, according to Yoon (), from which () below is drawn:

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  ()



kekcengha-koissta John-un Mary-ka o ci-anh-ul-ci John- Mary- come--- fear- ‘John fears that Mary might come (although it is unlikely to happen)’

However, there is no clear intuition that this interpretive effect can be generalized to French structures like (). There, there is no sense that the introduction of EN corresponds to the implicated meaning that the feared event is unlikely. Rather, the only detectable effect of EN on interpretation seems to be an effect of strengthening of the fear expressed, that is, a sort of emphatic effect with respect to the asserted meaning. Notice now that the semantic meaning of the implicated proposition (a) is more accurately represented as (c) below, and that (c) has in turn a pragmatic counterpart represented in (d): ()

c. It is likely and possibly certain that he comes d. It is likely but not certain that he comes

Suppose now that the operation of implicature denial encoded by EN does not apply in French to the first-level implicated proposition, that is (c), producing (b) (the correct result for Korean), but more locally to the second-level implicated proposition, that is (d), basically recovering and even strengthening the original semantic meaning in (c), as can be seen in (e): ()

e. It is likely and in fact certain that he comes

This would immediately explain the strengthening/emphatic effect detected in French (): the fear expressed is likely to be more robust if the feared event is virtually certain to happen. At the same time, what is going on here is still, in some sense, a process of ‘local’ implicature denial. Since this process applies to the second-order conversational implicature triggered by the implicated proposition, there is a strong resemblance with the locality effects detected with EN in comparative clauses: in comparatives, EN applies at the level of an embedded sentence, whereas with French verbs of fear it applies at the level of an embedded implicated proposition. We conclude that there is substantial evidence for the claim that EN is a polarity-reversal operator applying to implicated meaning and that this hypothesis may provide some new important avenues for future research.¹⁴

¹⁴ Of course, what I am suggesting is that a possible generalization of the Italian data and analysis to other languages might result in a valuable conceptual alternative to the expressive-evaluative analysis of EN, and result in an interesting research program for those who are inclined to believe that EN is not simply homophonous to real negation but constitutes a peculiar manifestation of real negation. Clearly, it is future comparative research that has to tell us to what extent implicature-denial can be generalized.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

.. C

.................................................................................................................................. In spite of the remarkable challenges it (still) poses, it is fair to say that the phenomenon of EN has already been elucidated along a number of syntactic and semantic dimensions. Syntactically, it has been argued that formal EN-licensing exhibits a significant intersection with NC-licensing (section ..).¹⁵ Semantically, it has been proposed that EN should be analyzed as a special sort of conventionalized implicature, triggering the activation of the evaluative layer of semantic meaning (section .). A different hypothesis, according to which EN, as a form of syntactically encoded negation, interacts with the layer of implicated meaning (as distinguished from the asserted meaning), has been explored in section .. and has been shown to give rise to many intriguing questions.¹⁶ Some of these questions are not entirely new. For instance, Krifka  offers a captivating analysis of a different sort of EN, exemplified in () above, which is licensed by before-clauses in some varieties of German, and involves double-negation structures, with one instance of negation in the main-clause (interpreted as a polarity-reversal operator) and another instance of negation in the embedded temporal clause (interpreted as a complement-set operator defined on times). Krifka interprets this instance of EN in German in incremental compositional terms, by making use of two-dimensional semantic representations (representing both the asserted and the implicated meaning). He shows that this incremental interpretive procedure makes either the assertion or the implicature informationally irrelevant, arguably producing the correct empirical results. From this perspective, it is worth noticing that the widespread use of EN that we have explored in the present chapter (in structures that usually involve only one instance of negation) suggests a non-incremental interpretation of negation, according to which negation dynamically interacts with the contextual determinants of implicated meaning. In fact, the very existence of EN as a syntactically encoded operator that reverses the polarity of some implicated proposition strongly suggests that syntax dynamically interacts with the enriched meaning that is created by perceptually and cognitively exploring the context in which the relevant syntactic structure is put to use. There is a potential parallelism to be drawn with the non-incremental models of negation in language processing, though there is no reason to believe that the ‘enriched’ semantics of negation envisaged here directly supports the existing models of non-incremental negation processing, such as the so-called “two-step simulation hypothesis” (Kaup, Zwaan, and Lüdtke ). What is clear, I believe, is that an enriched semantics of negation necessarily calls for a view of the syntax/semantics interface in which we do not simply compositionally interpret syntax, in a relatively context-independent way; rather, we apply at least some of the interpretive instructions encoded in the morpho-syntax of language to the representations produced in a much richer cognitive setting of which language is only a part.

¹⁵ Moreover, there is convincing evidence (which I have not reviewed here for reasons of space) to the effect that EN is a negative head filling a dedicated position within clausal structure, distinct from the position filled by real negation (Zanuttini and Portner ; Roguska ; Krifka ). ¹⁶ Some of these questions, which will not be addressed here for reasons of space, concern the status of metalinguistic negation (for a discussion, see especially Moeschler  and Delfitto, Melloni, and Vender ).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  ......................................................................................................................

                    Interaction of negation with quantifiers ......................................................................................................................

 

Whatever version of the grammar of negation is adopted, the analyst must eventually confront one of the most extensively studied and least-understood phenomena within the semantics of negation: the scope interaction of the negative operator with quantified subjects and with descriptions. (Horn : )

.. I

.................................................................................................................................. W an Amtrak train approaches a station with a short platform, passengers wishing to disembark must walk forward to a car that meets the platform. The conductor’s announcement to the passengers includes the statement in (). ()

All doors will not open.

U.S. coins come in denominations of  cent (penny),  cents (nickel),  cents (dime), and  cents (quarter). There is a children’s riddle that goes as in (). ()

Q: You have two coins that add up to  cents. One of them is not a quarter. What are they? A: A nickel and a quarter.

An August  interview on “Meet the Press” contained the following exchange between the U.S. President’s attorney (R. Giuliani) and the show’s host (C. Todd):¹

¹ Source:

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 ()

  : When you tell me that, you know, he should testify because he’s going to tell the truth and he shouldn’t worry, well that’s so silly because it’s somebody’s version of the truth. Not the truth. : Truth is truth. : No, no, it isn’t truth. Truth isn’t truth.

Examples () and () show what Carden () calls the - and - readings of negation, respectively. In () the passengers must (eventually) understand negation to take scope above the quantificational subject, all doors, yielding the negation of the proposition that all doors will open. In (), to solve the riddle one must take the negation in One of them is not a quarter to fall within the scope of the subject; as the answer to the riddle confirms, one of the coins indeed is not a quarter. (The other one is.) Carden’s terminology reflects the intuition that negation encompasses the quantifier in the one case (-) but only the verb phrase in the other (-). Finally, as () shows, a --like reading is possible even with a non-quantificational subject. Giuliani here does not claim that truth is equivalent to non-truth; rather, he rejects Todd’s tautological formulation, and by extension the notion that matters are quite so simple. In this chapter we examine the scopal interactions between negation and quantifiers. We will focus in particular on the question of what causes a quantificational subject to be understood as falling within the scope of clausemate sentential negation and vice versa, i.e. on how to understand and account for the distribution of the - and - readings. As Horn’s quote above attests, this is a puzzle with a long history. We begin by laying out some background assumptions about the syntax and semantics of scope-taking in section .. In section ., we examine a variety of recent proposals for how to account for the relative scope of negation and quantifiers. The accounts make appeal to phrase structure, focus, topic, and related pragmatic factors in order to explain why a particular environment might favor a particular scope relation. In section ., we examine some related issues in the grammar of quantifiers and negation. Section . concludes.

.. S       -

.................................................................................................................................. Quantificational DPs and negation are both standardly analyzed as scope-bearing expressions. While there is a wide variety of approaches to the semantics of scope, for present purposes I will assume that a scope-bearing expression denotes a function that takes a sentence denotation as its argument. Negation is typically analyzed as the truth-functional operator familiar from propositional logic; it flips the truth value of its argument. Quantificational DPs are typically analyzed as generalized quantifiers over individuals (after Barwise and Cooper ); they compose with an open sentence or, in the more commonly adopted style of implementation, with a predicate formed via abstraction over the free variable in an open sentence (see e.g. Heim and Kratzer ). In the generative tradition, scope relations are represented at the syntactic level of logical form (LF). Sentential negation is generally taken to be syntactically immobile, while quantificational DPs undergo (often obligatory) covert movement from their base/ θ-position to a scope position, in a process known as quantifier raising (QR). In the

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

    



implementation of May (), LF representations generated via QR do not necessarily disambiguate relative scope; more recent implementations, by contrast, generally take LF to encode scope relations unambiguously. In English and many other languages, sentential negation sits both linearly and hierarchically between the positions of the subject and the object.² An LF that preserves this hierarchical order will thus be one in which the subject takes scope over negation, and negation takes scope over the object (provided there is an available scope position for the object below negation, as indeed there must be if negation operates on sentences). As (a,b) show, quantificational objects tend to take scope below negation by default. The wide-scope reading for the object comes out more readily in (c), where the most salient reading is every > one > not. (For more on the role played by the dependent indefinite in bringing out this reading, see the discussion of Beghelli and Stowell () below in section ...) Positive polarity items like someone resist being in the scope of negation in examples like (a), where the only available reading is some > not. A narrow-scope reading is possible, however, in the presence of an appropriate higher operator, as in (b) (Baker ; Szabolcsi ). As a general matter, then, objects are able to take scope both above and below sentential negation. ()

a. John didn’t read every book. b. John didn’t tell everyone his secret. c. John didn’t tell everyone one piece of information.

()

a. John didn’t tell someone his secret. b. It’s a shame John didn’t tell someone his secret.

Subjects are likewise able in principle to take scope both above and below sentential negation: these are the relative scope configurations that yield the - and - readings, respectively. Unlike with objects, the phrase structure and syntactic derivation of subjects is usually taken to involve an interaction with the syntactic position of negation. Specifically, on many approaches the base/θ-position of the subject is within the (possibly extended; Kratzer ) projection of the verb, below negation, while its surface position is higher, in the tense projection above negation, where it moves for reasons of Case and/or EPP. The core syntax of subjects thus involves a movement/copying relationship that spans the position of negation, as sketched in (). ()

[TP Johni did [NegP not [vP ti read the book ]]]

We thus find an asymmetry between subjects and objects with respect to their scope relations with negation. An LF where an object takes scope above negation will be one where QR has moved it higher than necessary purely for purposes of interpretation; if QR is governed by Shortest Move, then this will always be an optional instance of QR (Takahashi ). An LF where a subject takes scope below negation, by contrast, will be one where quantifier lowering has taken place; in copy-theoretic terms, it will be an LF where the lower copy in an A-chain (including the determiner) is interpreted. In other words, it will be an instance of scope reconstruction. (For discussion of scope reconstruction in A-chains,

² Some languages make multiple positions available for negation. See Zanuttini () for a study of negation in Romance.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

see Boeckx ; for discussion of Scope Economy and related constraints on optional QR, and their interaction with negation, see Fox , Mayr and Spector , and Fleisher ; for discussion of more general constraints on QR, see Cechetto .) The relative scope of subjects and negation has received the bulk of the attention in the literature, and it will be our major focus in the remainder of the chapter. As we will see, authors differ on a number of basic questions in this area; one such issue is whether the unavailability of a certain scope configuration is due to the grammar not generating the corresponding LF or to some other principle filtering it out. Scope of negation is thus an area where we see a broader set of theoretical questions being debated: how much semantic and pragmatic information is represented directly in (and thus constrained by) syntactic structure, and how much should be attributed to extrasyntactic factors?

.. S 

..................................................................................................................................

... Scopal cartography It has long been recognized that unconstrained QR overgenerates readings in sentences with multiple scope-takers. Beyond asymmetries in the availability of inverse scope for different combinations of subject and object quantifiers, we can classify quantifiers according to their interactions with different classes of scope-takers. Beghelli and Stowell () develop a theory in which such constraints on relative scope receive not a semantic or a pragmatic explanation but a syntactic one, the result of an articulated clausal spine containing a variety of specialized functional heads. Constraints on scopal interactions between quantifiers and negation are, on this view, simply a special case of the more general limitations on scope-taking that follow from the exploded phrase structure of the clause. In Beghelli and Stowell’s system, quantifiers take scope not via ordinary QR but via feature-driven movement to the specifier of a functional head. The featural content of a given quantifier determines how high it may raise and thus where it may take scope: for some quantifiers, there is just a single functional head against which they may check their features, while other quantifiers have multiple options and a concomitant scopal mobility within the clause. Simplifying slightly, Beghelli and Stowell propose the following hierarchy of functional projections: ()

Ref > C > Dist > Share > Neg > V

Distributive universal DPs like those headed by each and every raise to the specifier of DistP, a relatively high position just below C, while negative DPs and sentential negation sit in NegP, a relatively low position. This phrase structure for the clause, in combination with the operative assumptions about how quantifiers take scope, predicts - readings for sentences with universally quantified subjects.³ As Beghelli and Stowell note, however, with

³ Beghelli and Stowell choose to focus on examples with neutral intonation. They do not offer an analysis of the - reading; they assume that the phonological characteristics typically needed to

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

    



neutral intonation even the ordinary - reading seems awkward and difficult to access; this is shown in ().⁴ ()

a. ?? Every boy didn’t leave. b. ?? Each boy didn’t leave.

Beghelli and Stowell observe that the acceptability of such configurations is greatly improved when a counting indefinite occurs in object position, as in (). The improvement, they note, is limited to the reading of such sentences in which the object scopes above negation, that is ‘for each boy x there is one book y such that x didn’t read y’. ()

a. Every boy didn’t read one book. b. Each boy didn’t read one book.

Beghelli and Stowell explain the contrast between () and () through appeal to the semantic needs of distributivity. The examples in () are awkward, they propose, because the distributive universal DP in SpecDistP lacks a distributed share in SpecShareP to distribute over. Counting indefinites like one book can check their features in SpecShareP; and since Share sits above Neg in the clausal spine, the result in () is an every > one > not scope configuration. The scope configuration every > not > one, with the indefinite below negation, would also be syntactically derivable (Beghelli and Stowell assume that counting indefinites can remain in their base positions). But such a configuration would yield the same semantic awkwardness as the ordinary - reading in (), and for the same reason: the distributive universal would lack a distributed share. The awkwardness of () and the strong preference for the every > one > not reading in ()—empirical data points that go overlooked in much of the literature on scope of negation—thus receive a common explanation in Beghelli and Stowell’s theory. Beghelli and Stowell go on to note an asymmetry in the ability of distributive universals to scope below negation when they occur in object position, as sketched in (). ()

a. John didn’t read every book. (not > every OK) b. John didn’t read each book. (each > not only)

They propose that universals headed by every optionally have a group-denoting interpretation that those headed by each lack. This enables every DPs to take scope below Neg when they occur in object position, whereas each DPs have no choice but to raise to DistP, and thus to outscope negation. See Beghelli and Stowell (: ff.) for discussion.

support the - reading lead to LFs different from the ones they propose for the intonationally neutral cases. ⁴ This fact is overlooked in many treatments of negation and quantifier scope. One exception is Steedman (: , ff.).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

... Negation and focus When focal emphasis is present, it can lead to ambiguities in the construal of negation. Such ambiguities can involve the apparent scope of negation, but they can also involve the pragmatic felicity of the sentence in different contexts and the range of felicitous continuations. Here we explore analyses that seek to account for such ambiguities while positing a consistent semantics for negation itself. Partee () discusses the example in (). As she notes, this sentence can be judged true in a context where Mary gave most or even all employees a raise, provided no raise was given on account of soft-heartedness. In other words, focal emphasis on soft-hearted yields a reading that is truth-conditionally distinct from the most salient reading of the sentence in the absence of such emphasis, which states that no employee got a raise and that the lack of raises was due (however counterintuitively) to Mary’s soft-heartedness. ()

Mary didn’t give any employee a raise because she was SOFT-HEARTED.

Partee pursues an analysis of sentential negation as an unselective quantifier (cf. Lewis ), that is as the operator in a tripartite structure. The role of focus is to help partition the sentence into restriction and nuclear scope for this operator, with the focused material mapped into the nuclear scope. The overall truth conditions for () then state that no instance of Mary giving an employee a raise was one in which she gave that employee a raise because she was soft-hearted; put differently, for any instance of Mary giving an employee a raise, it is not the case that Mary gave that employee a raise because she was soft-hearted. As this paraphrase suggests and as Partee discusses, this allows us to treat the nuclear scope as the scope of negation, while integrating the apparent focus-sensitivity of negation into a more generally applicable framework for focus and quantification. (Partee does not take the licensing of any in the restriction here to warrant any particular conclusion about the restriction’s falling in the semantic scope of negation; as she notes, NPI any is licensed quite broadly in operator restrictions.) In a similar vein, Herburger () develops an analysis in which the different construals of negation in the presence of focus are tied to the partition of the sentence into restriction and nuclear scope. In Herburger’s neo-Davidsonian analysis, however, the operator is not negation itself but an event quantifier, and the ambiguities involving negation result from negation’s being mapped to different positions in the articulated logical structure. For the example in (), Herburger proposes that negation can apply to the entire nuclear scope as in (a), to the verb visit as in (b), or to the entire proposition as in (c). (I omit tense and contextual restrictions from these logical forms.) ()

John didn’t visit MONTMARTRE. a. 9e [visit(e) & Ag(J,e)] ¬[Th(M,e) & visit(e) & Ag(J,e)] b. 9e [¬visit(e) & Ag(J,e)] [Th(M,e) & ¬visit(e) & Ag(J,e)] c. ¬9e [visit(e) & Ag(J,e)] [Th(M,e) & visit(e) & Ag(J,e)]

Roughly speaking, (a) says that there was a visiting event whose agent was John, but that Montmartre was not its theme; (b) says that there was an event whose agent was John and whose theme was Montmartre but which was not a visiting event; and (c) says that there

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

    



was no visiting event whose agent was John and whose theme was Montmartre. Since existential quantification is semantically symmetrical, the partitioning into restriction and nuclear scope does not affect the sentence’s truth conditions in any of these cases. Rather, Herburger proposes that the role of the restriction is to establish what she calls a “backgrounded focal entailment,” essentially an aboutness relation. With a restriction as in (a, c), the sentence is about events of visiting whose agent is John; with the restriction in (b), the sentence is about events whose agent is John which are not visiting events. Each example is pragmatically appropriate in a context that makes its restriction salient. The highly articulated structure makes available a variety of scope positions for negation—in each case modeled as a propositional operator—with consequences for interpretation within the pragmatics rather than the truth-conditional semantics. It is a bit less clear how Herburger’s system would handle examples in which negation and quantifiers interact. We could in principle generate at least six different logical forms for an example like (). ()

John didn’t read EVERY BOOK. a. 9e [read(e) & Ag(J,e)] ¬[every-book[x] : Th(x,e) & read(e) & Ag(J,e)] b. 9e [¬read(e) & Ag(J,e)] [every-book[x] : Th(x,e) & ¬read(e) & Ag(J,e)] c. ¬9e [read(e) & Ag(J,e)] [every-book[x] : Th(x,e) & read(e) & Ag(J,e)] d. every-book[x] : 9e [read(e) & Ag(J,e)] ¬[Th(x,e) & read(e) & Ag(J,e)] e. ¬every-book[x] : [9e [read(e) & Ag(J,e)] [Th(x,e) & read(e) & Ag(J,e)]] f. [every-book[x] : ¬9e [read(e) & Ag(J,e)] [Th(x,e) & read(e) & Ag(J,e)]]

The logical forms in (a, b, c) are parallel to their counterparts in (), with the object quantifier scoping within the nuclear scope of the event quantifier. Each of these three logical forms yields truth conditions that are too weak for either the not > every or the every > not reading of the sentence; each is compatible with a scenario in which John did in fact read every book (provided this outcome can be achieved over a collection of distinct events, in the case of (c)). The logical form in (d) is likewise too weak, requiring of each book merely that there be some event where John read something other than it.⁵ The not > every and every > not readings are captured by (e) and (f ), respectively, both logical forms where negation scopes above the event quantifier. The multiplicity of available scope positions thus may overpredict the range of readings found when negation interacts with a quantifier.

... Negation and topic It is widely observed that the - reading requires—or at least strongly prefers—a particular variety of discourse context and a marked intonation. In the generative tradition, there are a variety of theories of how intonation interacts with scope; a foundational early work is Jackendoff (). In this section, we explore the approach of Büring (a, b), who develops a formal theory of topic and focus and shows how it can illuminate our understanding of quantifier–negation interactions. ⁵ An anonymous reviewer suggests that this reading may indeed be available, characterizing a scenario where each book was such that John avoided reading it on some occasion or other (though he may ultimately have read all of them).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

Büring sets out to explain how intonation influences the calculation of scope. His particular focus is on sentences with a universal subject headed by all (not every or each) alongside sentential negation. Not being obligatorily distributive, all supports a - reading with neutral intonation or with an A accent (Bolinger ; Jackendoff ), where emphasis falls on all. By contrast, when there is a second phonological emphasis on negation, or when the sentence is intoned with the B accent, where there is a phrase-final rise in pitch, the - reading becomes strongly preferred. The contrast is sketched in (). ()

a. ALL politicians are not corrupt. (A accent: -) b. ALL politicians are NOT corrupt. (B accent: -)

Büring assumes that the operative syntactic principles of scope assignment permit both relative scopings in (), that is that the grammar generates LFs for both the - (all > not) and the - (not > all) readings. Focusing on the case of (b), he argues that disambiguation in favor of - is the result of a formal pragmatic requirement that filters out the - LF due to its failure to generate what he calls a Residual Topic. Relative scope, then, is limited in this case not by any grammatical constraint on the derivation of LFs, but by a (suitably formalized) discourse coherence condition. Büring assumes the presence of Focus and Topic features at LF that enter the recursive semantics. Büring adopts Rooth’s (, ) alternative semantics for focus, where ordinary semantic values exist alongside focus-semantic values, which are sets of ordinary semantic values (e.g. sets of propositions). Büring treats topic as a kind of higher-order focus: topic-semantic values are sets of focus-semantic values (e.g. sets of sets of propositions). On a Hamblin/Karttunen semantics for questions, this makes the focus-semantic value of a sentence equivalent to a question, and the topic-semantic value of a sentence equivalent to a set of questions. In Büring’s analysis of (b), there is a Focus feature attached to not and a Topic feature attached to all. This means that focus-semantic values will show variation in the position of not: a focus-semantic value will be a two-membered set consisting of the negated proposition in question and its affirmative counterpart. Topic-semantic values will show variation in the position of the quantificational determiner: each element will be a focussemantic value as just described with some determiner or other in the position of all. When all scopes above not, as on the - reading, we get the topic-semantic value shown in (). ()

Topic-semantic value for all > not: {{all(politicians)(λx.¬corrupt(x)), all(politicians)(λx.corrupt(x))}, {most(politicians)(λx.¬corrupt(x)), most(politicians)(λx.corrupt(x))}, {some(politicians)(λx.¬corrupt(x)), some(politicians)(λx.corrupt(x))}, . . . }

Büring’s discourse coherence condition states that the utterance of a Topic-containing sentence in a given context must leave at least one element of the sentence’s topic-semantic value unsettled in the updated context. Put differently, the condition states that the utterance must leave at least one of the questions that constitute the sentence’s topicsemantic value unanswered, or “disputable.” The disputable element(s)/question(s)

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

    



remaining form what Büring calls the Residual Topic. Put as succinctly as possible, then, Büring’s condition states that the utterance of a Topic-containing sentence in a given context must leave behind a Residual Topic. Büring attributes the unavailability of the - reading in (b) to its failure to leave behind a Residual Topic. The assertion, with all scoping over not, resolves the first question in the topic-semantic value shown in (): there is no longer any doubt as to which of the two propositions all(politicians)(λx.¬corrupt(x)) and all(politicians)(λx.corrupt(x)) is true in the updated context. The problem is that, having asserted the truth of all(politicians)(λx. ¬corrupt(x)), there is no longer any doubt as to which proposition is true in any of the other questions that make up this topic-semantic value. There is thus no Residual Topic, and the sentence is infelicitous on this reading. The topic-semantic value for the - reading, by contrast, where not scopes above all at LF, is as in (). ()

Topic-semantic value for not > all: {{¬all(politicians)(corrupt), all(politicians)(corrupt)}, {¬most(politicians)(corrupt), most(politicians)(corrupt)}, {¬some(politicians)(corrupt), some(politicians)(corrupt)}, . . . }

Here uttering the sentence adds the proposition ¬all(politicians)(corrupt) to the common ground. This once again resolves the first question in the topic-semantic value, but it leaves each remaining question unresolved: if it is merely the case that not all politicians are corrupt, then it may be the case that most politicians are corrupt, or not; and so on for the other remaining questions. On this reading, then, there is a Residual Topic, and Büring’s discourse coherence condition does not filter out the underlying LF as it does in the case of the - reading. To summarize: for Büring, the effect of intonation on the scope of negation is mediated by (i) the syntactic Focus and Topic features associated with particular intonational melodies, (ii) the semantic effects of those features within an expanded Roothian alternative semantics, and (iii) conditions on discourse coherence stated within that formal framework. Within this approach, scopal asymmetries reflect the filtering effect of discourse conditions, not lower-level constraints on the generation of logical forms. For further work within this approach, see Büring () and Wagner (, ).

... Scope of negation in Horn’s Extended Term Logic Horn (), in the final chapter of his classic study of natural-language negation, develops a version of Aristotelian term logic he dubs Extended Term Logic (ETL). Horn’s account of scope of negation is couched within this system, whose mechanisms for introducing negation and for permitting scopal interactions between negation and quantifiers differ in important ways from the generative systems discussed above. Within ETL there are two distinct processes that introduce negation: predicate term negation and predicate denial. Predicate term negation applies to a predicate and yields a new predicate that can compose with a subject; its semantics is that of contrary opposition.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

Predicate denial is a mode of composition that combines a subject and a predicate to create a proposition; its semantics is that of contradictory opposition.⁶ ETL does not countenance the truth-functional negation operator familiar from propositional logic and most generative treatments. ETL’s two processes for introducing negation correspond straightforwardly to the two scope interpretations for negation: predicate term negation negates the predicate and thus yields a - reading, while predicate denial denies that a predicate holds of a (possibly quantificational) subject, yielding a - reading. Horn’s discussion focuses on cases that appear to flout the systematicity of this relationship between mode of composition and semantic interpretation; in particular, Horn is concerned with cases where predicate denial yields a - reading. Horn’s account of why a predicate denial syntax is so often paired with a - semantics makes appeal to a confluence of functional tendencies and pressures. First, Horn notes that syntactic subjects are also frequently pragmatic topics, and that topics are given or established in the discourse in a way that can conflict with negation’s taking scope above them. This is true in particular for names and definite descriptions, whose existential presuppositions are often preserved under predicate denial. In this connection, Horn cites Kuroda’s () distinction between thetic and categorical judgments, work with roots in the Prague School. In a categorical judgment, an established topic is taken up and then a predication is made of it; to deny that a predicate holds of an established topic, in this approach, is not to deny the semantic and pragmatic import of the topical phrase itself. The pragmatic privileges of topichood are also available in principle to quantificational subjects, and as Horn notes, predicate denial often leads to a narrow-scope, - reading of negation with quantificational subjects. Of particular interest here is an asymmetry between universally and existentially quantified subjects. As shown in (), universals appear to support the wide-scope, - reading much more readily than existentials do. ()

a. Everybody didn’t come. b. Somebody didn’t come.

To the various functional pressures Horn cites as favoring the narrow-scope, - reading of negation in these predicate denials, he adds one more: competition with a lexicalized alternative, combined with the cross-linguistic tendency for languages to lexicalize negated existential (not > some) operators but not negated universal (not > every) ones. The - reading of the predicate denial in (b) can be unambiguously expressed by composing the predicate (came) with a lexicalized negative existential quantifier (nobody). The corresponding construction in the case of (a), by contrast, requires the use of a morphologically marked phrasal constituent (not everybody). Horn appeals to this difference in relative markedness to explain the more ready availability of the - reading in (a).

⁶ Contrary opposites cannot be simultaneously true, though they may be simultaneously false; Everybody came and Nobody came are contrary opposites. Contradictory opposites cannot share truth value, which means that exactly one of them is true and the other is false; Somebody came and Nobody came are contradictory opposites.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

    



As Horn (: ff.) notes, when the pragmatic effects of topichood are controlled for or overridden by other discourse-pragmatic pressures, the - reading of predicate denial can more easily emerge, regardless of the quantificational force of the subject. Horn thus does not envision a strict mapping between syntactic mode of composition and semantic scope of negation. Rather, the semantic interactions between negation and quantifiers are but one case of a broader pattern, one where pragmatic factors are a driving force in interpretation rather than a filter on grammatical outputs.

.. O 

..................................................................................................................................

... Scope of incorporated and covert negation Beyond its role as a sentential operator, negation has been argued to be a component of certain quantificational elements and phrasal constructions. Evidence in favor of this view comes from the apparent scopal independence of the negative component in such constructions. Quantifiers and constructions that have been analyzed in this way include negative existentials (also known as negative indefinites), the comparative degree operators less and fewer, and comparative than clauses. The component negation analysis of negative existentials offers an account of the scope splitting found in examples like (), with the German negative existential kein (Jacobs ), and (), with English no (Iatridou and Sichel ). ()

Alle Ärzte haben kein Auto. (¬ > alle > 9)

()

No recording of this session may be made. (¬ > may > 9)

In these examples, a scope-bearing element—the universal quantifier alle Ärzte or the modal may—takes scope between the negative and existential components of the negative existential quantifier (kein Auto or no recording). While the negative component is scopally independent of the existential component, the relationship between these two elements is not unrestricted: though another operator or operators may take scope between them, the negative component always scopes above the existential component. Jacobs (: ) notes that the split-scope reading of () is most readily available with topical emphasis on alle and focal emphasis on kein, precisely the intonational pattern that demands a - reading in the examples discussed above analyzed by Büring (a, b). The comparative degree operators fewer and less (which may be analyzed as count and mass allomorphs, respectively, of the same item) are taken by Heim () to consist of the positive degree operator more plus a scopally superior negation. Evidence comes once again from scope splitting, as in the following examples (see also Hackl ). ()

a. You have to be less tall than John to ride this roller coaster; you only have to be as tall as Mary.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



  b. You have to answer fewer than  questions to get full credit; you only have to answer .

As the continuations above make clear, the sentences in () have an available reading where the modal have to takes scope above the degree-inequality component of the degree operator but below the negative component: a split-scope reading (schematically, ¬ > have to > more). Penka (: ch. ) catalogues a number of other constructions that support split-scope readings of this sort, including only and modified numerals with at most. Finally, there is a long-standing analysis of clausal comparatives, originating with Seuren (), according to which the comparative than clause contains a covert negation. (More recent exponents include Gajewski  and Schwarzschild .) Known as the “A-not-A” analysis, it has been most recently updated by Alrenga and Kennedy (), who propose that the negative element in the than clause is not a covert sentential negation but a covert negative degree operator. This covert negative’s scope is then limited by known constraints on degree-operator scope; in particular, it may not take scope above a subject quantifier within the than clause.

... Negative islands Scopal interactions between negation and interrogative operators are limited in interesting ways. In particular, when negation occurs between a degree interrogative phrase and its trace, the result is often infelicitous, as in (). Such configurations have come to be known as negative islands. An appropriately placed modal can have a rescuing effect, as in () (Fox and Hackl ; Abrusán and Spector ). ()

a. How fast was John driving? b. #How fast wasn’t John driving?

() a. How fast is one not allowed to drive? b. How fast is one required not to drive? Cinque () and Rizzi () propose syntactic accounts that attribute the negative island effect to a violation of Relativized Minimality (in line with other weak island effects). By contrast, Szabolcsi and Zwarts () place the problem squarely in the denotational semantics. On their account, a degree interrogative quantifies over an expression with the semantic structure of a join semilattice; negation effects Boolean complementation, an operation not defined on such domains. In a similar vein, Rullmann (b) proposes that the negative island effect is the result of a semantic crash in which the maximalization operator introduced by the degree interrogative is applied to an argument that lacks a scalar maximum. Beck and Rullmann () propose an amended semantic account that attributes the negative island effect to the failure of a maximal informativity presupposition, that is a presupposition that the degree question has a strongest true answer (after Dayal ). This account affords better empirical coverage, as the negative island effect is tied not to the matter of scalar maxima or minima but to the inferential relationships among propositions

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

    



that include particular scalar values. Fox and Hackl () expand on this to argue in favor of the view that scales are dense, showing that the negative island effect arises when we seek a maximally informative answer to a question about degrees above or below (but not equal to) a certain value. See Abrusán and Spector () and Abrusán () for additional work in this vein.

.. S

.................................................................................................................................. The scopal interactions between negation and quantifiers are at once straightforward to describe and difficult to explain. The core question of under what conditions a quantificational subject will be interpreted as falling within the scope of sentential negation and vice versa is one that has engendered a large literature; space has precluded an examination of anything more than a small portion of it here. Factors as varied as the syntax of subjecthood, the phrase structure of scope taking, the pragmatic partition of the sentence into focus and background, and the discourse conditions regulating topichood have been argued to play a role, directly or indirectly, in regulating these scopal interactions. Broader questions about the proper division of labor between grammatical and extragrammatical factors, and about the line of demarcation between the two, are at play in the variety of approaches considered here. The relative scope of negation and quantifiers remains an active area of investigation still in search of a unifying account.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  ........................................................................................................................

SEMANTICS AND PRAGMATICS OF NEGATION ........................................................................................................................

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  ......................................................................................................................

   ......................................................................................................................

    

. I

.................................................................................................................................. I many languages, modals have fixed scope with respect to negative expressions.¹ For example, English modals must, should, ought to, be to, might, and epistemic may obligatorily scope above negation, whereas need, can, have to, and deontic may obligatorily scope below it (Cormack and Smith ; Butler ); this is demonstrated in () and (). Similar facts have been reported for Dutch, German, Greek, and Hindi (von Fintel and Iatridou ; Iatridou and Zeijlstra ), Catalan (Picallo ), and French (Homer , ). ()

()

Deontic a. Diana mustn’t/shouldn’t/isn’t to leave. b. Diana needn’t/can’t2/may not/doesn’t have to leave. Epistemic a. Agatha mustn’t/shouldn’t/may not/might not have left. b. Agatha can’t/doesn’t have to have left.

 > , * >  * > ,  > 

 > , * >  * > ,  > 

This pattern is curious for two reasons. First, the scope of modals in English (though not in all languages)³ does not correlate straightforwardly with modal strength or with modal

¹ There will be several caveats on this point, to be revealed shortly. For now, we consider only sentences without special/marked prosody. ² A reviewer correctly points out that negation can scope under deontic can under certain conditions. The same is true of deontic may. i. She can/may [not go to school (if she feels unwell)].

/ > 

This may well be a case of constituent negation, as evidenced by the fact, also pointed out by the reviewer, that this negation cannot contract. ³ Picallo () reports that in Catalan, epistemic modals scope above negation while deontic modals scope below it.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



    

flavour.⁴ There are epistemic modals that scope above negation and epistemic modals that scope below it; there are deontic modals that scope above it and deontic modals that scope below it; there are universal modals that scope above it and universal modals that scope below it; and there are existential modals that scope above it and existential modals that scope below it. Second, semantic scope of modals need not be reflected by the surface scope of the operators involved; for example, although must and need both appear above negation, the former scopes above negation while the latter scopes below it. Similar mismatches between surface scope and semantic scope have been observed for deontic modals in German, Greek, and Hindi (Iatridou and Zeijlstra ). ()

German a. Hans muss nicht abfahren. Hans must  leave ‘Hans doesn’t have to leave.’ b. . . . dass Hans nicht abfahren muss. that Hans  leave must ‘ . . . that Hans doesn’t have to leave.’

()

Greek a. Dhen chriazete na figis.  need  leave ‘You don’t need to leave.’ b. Dhen prepi na to kanume afto.  must  it do this ‘We must not do this.’

()

Hindi a. tumhen Dilli nahiiN jaa-naa hai. you. Delhi  go- be. ‘You don’t have to go to Delhi.’ b. tumhen Dilli nahiiN jaa-naa caahiye. you. Delhi  go- should ‘You should not go to Delhi.’

 > , * > 

 > , * >  (Iatridou and Zeijlstra : )

 > , * > 

* > ,  >  (Iatridou and Zeijlstra : )

 > , * > 

* > ,  >  (von Fintel and Iatridou : )

These facts demand an explanation. How do modals in these languages achieve rigid and moreover possibly non-surface scope with respect to negation? Why does relative scope with respect to negation differ between these modals, even those that appear to surface in one and the same position? This chapter will focus on three ways of deriving the facts that have been proposed in the literature: (i) different positions of interpretation, (ii) polarity sensitivity, (iii) Neg-Raising.

⁴ By “strength” we mean here the quantificational force of the modal, and by “flavour” we mean the combination of modal base and ordering source in the sense of Kratzer (, ).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



.. D   

.................................................................................................................................. One way of capturing the rigid scope of modals with respect to sentential negation has been to posit different positions of interpretation for different modals (just as Beghelli and Stowell  propose for quantificational DPs). Some of these positions happen to be above negation while others happen to be below it. For example, Cormack and Smith () propose that there are two LF positions for modal auxiliaries in English, flanking the position where sentential negation is interpreted.⁵ ()

C T (Modal1) Pol(/) (Modal2) . . . (adapted from Cormack and Smith : )

Which modals occupy which position is a matter of lexical specification on this approach; some modals select for a Polarity (Pol) phrase at LF and so are inserted and interpreted above negation while others are not. Cormack and Smith make the unusual move of merging the LF- and PF-interpretable parts of a sign in potentially different positions, bypassing the puzzle of why modals can be pronounced in one position and interpreted in another. Butler () proposes a similar system with four positions for English modals: one each for epistemic necessity, epistemic possibility, root necessity, and root possibility. He assumes a different set of facts from what was laid out in section .; more particularly, he assumes that (i) epistemic modals always scope higher than root modals (cf. Cinque , i.a.), and (ii) necessity modals always scope above negation while possibility modals always scope below negation, as in ().⁶

⁵ Cormack and Smith note that sentential negation must be distinguished from both metalinguistic/ denial negation and VP constituent negation; all modals scope below metalinguistic/denial negation and above VP negation, but they differ in their scope with respect to sentential negation. Cormack and Smith propose distinct positions for the three types of negation, with VP negation located structurally below Modal2 and metalinguistic/denial negation above Modal1. ⁶ This is accurate for deontic modals if we ignore the somewhat marked form need and semi-modals like have to, but Butler () states that epistemic possibility modals likewise all take scope below negation; while this is correct for can, it is not correct for might and may. His system allows the full range of data to be captured if we assume that may and might take scope above the lower negation (see Butler :  for the two positions of negation). i. [ForceP [FocP neg [FinP canepi [TP subject [T’ T [ForceP [FocP [FinP vP]]]]]]]]

 > 

ii. [ForceP [FocP [FinP might/mayepi [TP subject [T’ T [ForceP [FocP neg [FinP vP]]]]]]]] / >  However, the question then arises why the following configuration, which would allow can to take scope over negation, is not available: iii. [ForceP [FocP [FinP canepi [TP subject [T’ T [ForceP [FocP neg [FinP vP]]]]]]]] Butler’s theory is silent on this matter.

* > 

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 ()

     epistemic necessity > (negation) > epistemic possibility > (strong) subject > root necessity > negation > root possibility > vP (Butler : )

Butler () argues that negation can act on any propositional constituent, and so proposes that there are two distinct positions for negation, one at the edge of vP and one at the level of CP. He proposes that the higher negation is flanked by the two positions for epistemic modals, and the lower negation is flanked by the two positions for deontic modals, as in (). ()

LF: [ForceP necepi [FocP neg [FinP possepi [TP subject [T’ T [ForceP necdeo [FocP neg [FinP possdeo vP]]]]]]]] (adapted from Butler : )

Butler () suggests that modals are merged low as verbal projections and get to their specified LF positions via covert movement or some equivalent mechanism, although he does not provide specifics of how this movement operates. These accounts are a sort of “just-so” story: they might be able to in principle accurately capture the data that they set out to derive, but they do not attempt to explain why the scope of modals is the way it is. On Butler’s view, it is not clear why possibility modals should be interpreted in a different position from necessity modals, and why the former position is lower than the latter, for both epistemic and for root modals (including deontics). On Cormack and Smith’s view, it is not clear why some modals select for Pol and others not. One could argue that “just-so” stories is all there can be in the study of interaction of modals and negation. That is, that there is no deeper truth to the matter. After all, there are some such facts in the natural world, of which language is a part. But in general, it is hard to distinguish between the validity of “justso” stories and their insufficiency. So even if, in principle, there may not be a deeper truth to the matter, the literature has attempted to find one. We turn to some such accounts in the remainder of the chapter.

.. P 

.................................................................................................................................. There is a strand of literature that argues that certain modals take rigid scope with respect to negation because they are polarity-sensitive. On this view, regardless of where they appear in the surface string, some modals must scope over negation because they are PPIs (Positive Polarity Items); others must scope under it because they are NPIs (Negative Polarity Items). Such accounts aim to capture both the particular scopes that are available to modals and the rigidity of these scope possibilities. This view is motivated by a desire to capture additional sets of data not covered by the positional accounts. We describe these additional phenomena in turn.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



... NPI modals There are differences among the modals that scope below negation: some, like English deontic need and epistemic can, Dutch hoeven, and German brauchen,⁷ do not just scope under negation but, in fact, require the presence of negation: ()

a. You need*(n’t) say more. b. She can*(’t) have left. (It’s very early) c. Je hoeft *(niet) weg te gaan you need *(not) away to go d. Hans braucht *(nicht) zu gehen Hans need *(not) to go

Other modals, like English deontic can, have to, and German müssen, scope under negation but do not require its presence. ()

a. You can(’t) leave. b. You (don’t) have to leave.

For the modals that require a negative environment, this requirement can be satisfied by expressions other than sentential negation; they are also acceptable in the scope of negative DPs (henceforth: NegDPs) and only. One salient property that unifies these environments is that they all license NPIs. It has therefore been argued that modals like English deontic need (and epistemic can, Dutch hoeven, and German brauchen) are NPIs (Edmondson ; van der Wouden ; Israel , i.a.), while modals like English deontic can and have to are not polarity-sensitive. ()

a. You needn’t worry. b. No one need leave. c. Only linguists need apply.

... PPI modals There are modals that scope over negation, no matter where they appear in the surface string with respect to negation. This raises the possibility that these modals may be PPIs. On its own, wide scope with respect to negation does not suffice to make an item a PPI, and

⁷ Hoeksema () claims that need, hoeven, and brauchen are also licensed in questions. In contrast, Lin, Weerman, and Zeijlstra () claim that hoeven is only acceptable in environments that can be decomposed to contain a negation (i.e. not in restrictors of universal quantifiers, antecedents of conditionals, questions, etc.). English need is acceptable in some questions (e.g. Need I say more?), but this construction is not very productive; then again, auxiliary need is in general on its way out in North American English. Need is acceptable under hardly, which is probably plain downward-entailing. Israel () reports that epistemic can is licensed in questions.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



    

so a number of researchers have looked for and found properties that align these modals more generally with behavior associated with PPIs (for general PPI behavior, see Szabolcsi ). More specifically, there are circumstances where modals that ordinarily take scope above negation can take scope below it (Iatridou and Zeijlstra , ; Homer ; see also Cormack and Smith ), a fact that is not predicted by the positional accounts described in section .. These all happen to be environments where PPIs are known to be able to take scope under negation (Szabolcsi ). This fact has led Iatridou and Zeijlstra (, ) and Homer (, ) to argue that these modals are PPIs (see also Israel ). Let us consider the evidence more closely. Modals like must and should take scope below negative material when it is interpreted as metalinguistic or contrastive (Iatridou and Zeijlstra ; see also Cormack and Smith ). These interpretations require a special context, and often special non-default prosody (see Horn ,  on this point for NPIs in general, not modals). ()

No one MUST leave—everyone can do what they like!

()

A: Nadia should leave. B: No she shouldn’t!

 > 

 > 

This kind of negation does not anti-license PPIs, just as it does not license NPIs (Karttunen and Peters ; Horn , ; Szabolcsi ). ()

a. Sam didn’t buy SOMETHING—she bought EVERYTHING! b. *Sam didn’t buy ANYTHING—she bought EVERYTHING!

 > 

Moreover, modals like must can take scope below negation when an appropriate operator intervenes; this is illustrated in () with the universal quantifier everyone (Iatridou and Zeijlstra ; Homer ). ()

Not everyone must leave.

 >  > 

This configuration is dubbed shielding by Szabolcsi (); it disrupts polarity item (anti-) licensing for non-modals, as illustrated in () for the universal quantifier always (cf. Linebarger’s () Immediate Scope constraint). ()

a. Chris doesn’t always buy something. b. *Chris doesn’t always buy anything.

 >  >  * >  > 

Finally, modals like must can take immediate scope below negative material when the negation-modal combination is itself embedded in an NPI-licensing environment (Iatridou and Zeijlstra ; Homer ). ()

a. Emilia didn’t say that no one must leave.  >  > ,  > >  b. If no one must leave, we will stay here all day.  >  > ,  >  > 

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



This configuration, which Szabolcsi () calls rescuing, is known to have the same effect on non-modal PPIs (Baker ; see also Jespersen –).⁸ ()

a. I don’t think that Hestia didn’t say something. b. If we don’t call someone, we are doomed.

 >  > ,  >  >   >  > ,  >  >  (Szabolcsi : , )

... Neutral modals Based on the facts discussed in sections .. and .., Israel (), Iatridou and Zeijlstra (, ), and Homer (, ) propose that, in addition to there being modal NPIs, there are modal PPIs. An account in terms of polarity sensitivity provides a straightforward reason for some modals to take rigid scope above negation and others below it; their scope with respect to negation follows from their licensing needs. However, even if some modals are polarity items (some NPIs, some PPIs), not all modals fall into one of these categories. This means that even the most successful polarity account cannot capture the behavior of what Iatridou and Zeijlstra () call “neutral” modals without saying something more. English modal have to and deontic can are neutral modals in that they do not require a negative environment, yet scope under negative material when such material is present: ()

a. Dorcas has to leave b. Dorcas does not have to leave c. Nobody has to leave

()

a. Philomena can leave b. Philomena can’t leave c. Nobody can leave

 >  , *  >   >  , *  > 

 > , * >   > , * > 

As the positive sentences in (a) and (a) are both acceptable, the narrow scope of these modals in (b–c) and (b–c) cannot be attributed to NPI licensing conditions. To derive the rigid scope of non-polarity-sensitive modals with respect to negative material when such material is present, Iatridou and Zeijlstra () propose that all English modals start out below negation and those that can move (like auxiliaries can, must, and so on, but not have to), move overtly to I⁰ because they are tensed.⁹ Their

⁸ Jespersen noted rescuing for PPIs in cases where the second anti-licenser is negation and proposed that the two negations cancel each other out. Baker noted that other NPI licensers can play the role of the second negation. ⁹ This is contra Chomsky () and much subsequent work in which English modal verbs appearing in I⁰ are base-generated there. Iatridou and Zeijlstra () argue that this common position is undetermined by the data, and, in fact, wrong. In Chomsky (), modals in I⁰ differ from other verbs in that (a) they only have tensed forms (no infinitives or participles) and (b) they are generated in I⁰. However, Iatridou and Zeijlstra argue that the first of these two properties derives the second one: these modals are

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



    

position is that this head movement obligatorily reconstructs unless reconstruction would produce a polarity item licensing violation. This, they argue, is why polarity-neutral modal auxiliaries take scope below negation even when they appear above it in the surface string, like deontic can. Because it is not polarity-sensitive, reconstruction would not lead to a polarity item licensing violation, and so can reconstructs and is interpreted below negation. Other polarity-neutral modals like have to likewise take scope below negation because they never move above it in the first place. Similarly, NPI modals must reconstruct, because doing so would not yield a polarity licensing violation—on the contrary, failure to reconstruct would result in the NPI modal being interpreted outside of the scope of its licenser. Their obligatory narrow scope with respect to negation follows. PPI modals, however, are banned from reconstructing across negation, because to do so would result in them being interpreted in the immediate scope of an anti-licenser; thus, they take wide scope over sentential negation. According to Iatridou and Zeijlstra (), then, the English modal system divides up as in Table ..

Table . English¹⁰ modals and polarity sensitivity Deontic

PPI¹¹

Neutral

must, should, ought to, be to

can, may, have to, need to need

Epistemic¹² must, should, ought to, may, might could, have to

NPI

can

movers (they move to C⁰ in questions: Can/may/must he leave?) and since they only exist in tensed forms, they always have to move to I⁰ (in English, tensed verbs that are able to move, always move to I or higher). So, according to Iatridou and Zeijlstra, the position that modals that appear in I are generated lower than I is as consistent with the facts as the generation-in-I⁰ view is. ¹⁰ This table reflects North American English modals. Cormack and Smith () report that both epistemic and deontic need (cf. need to) obligatorily scope below negation in UK English, but Butler () notes that epistemic need is all but non-existent in the form of English he describes. ¹¹ According to Iatridou and Zeijlstra (), modal PPIs, like DP PPIs, come in different strengths. ¹² The argument for NPI and neutral status proceeds for epistemic modals much as it does for deontics and so is omitted here in the interest of space. The argument for PPI status, however, goes a bit differently. For deontic modals, two facts ruled out a positional analysis whereby some modals scope above sentential negation simply because they are generated above it: these modals (a) take scope above subject NegDPs and (b) can take scope below negative material in exactly those situations where PPIs are generally known to do so (contrastive/metalinguistic contexts, rescuing, and shielding). Concerning (a), at least some epistemic modals that take wide scope with respect to sentential negation (e.g. must) take wide scope with respect to subject NegDPs. However, because the Epistemic Containment Principle (von Fintel and Iatridou ) independently forces epistemic modals to take scope over quantifiers (like NegDP subjects) that have moved across them, this does not constitute a clear argument for PPI status. Moreover, there is variation among the epistemic modals that take scope above negation in this respect; epistemic must easily takes scope above subject NegDPs, but may and might have difficulty doing so.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



Modals that scope in a particular way with respect to sentential negation scope in the same way with respect to other negative expressions, such as (the negative portion of ) NegDPs (Iatridou and Sichel ): ()

a. No one must leave. b. No one has to leave.

 > ; * >  *  > ;  >  

... Further challenges Even if the polarity-(in-)sensitivity approach is correct, there are still parts of the data that are not covered. In order to avoid being anti-licensed, PPIs that appear in head-initial languages to the left of (i.e. above) negation in the surface string simply do not reconstruct, as mentioned above. However, there are cases where the anti-licenser is above a PPI modal in the surface string, and so not reconstructing is not enough to save the PPI from being anti-licensed: ()

No one must leave.

* > ,  > 

()

Dhen prepi na to kanume afto.  must  it do this ‘We must not do this.’

* > ,  > 

i. You knock on your friend’s door and ring the doorbell but there is no answer. a. No one must be home. * > ,  >  b. ??No one may/might be home. * > /, ??/ >  As for (b), the Epistemic Containment Principle makes shielding configurations difficult to test; because shielding requires the alleged PPI to take scope below an anti-licenser and an intervening quantificational element, and because the only anti-licensers that appear above modals are NegDP subjects, this configuration will require that an epistemic modal take scope below a quantifier that has moved across it—exactly the configuration that is banned by the Epistemic Containment Principle. Epistemic Containment likewise complicates the rescuing diagnostic for PPIs. In the demonstration of rescuing for deontic PPI modals, the immediate anti-licensing environment was a NegDP. It turns out that narrow scope for deontic PPI modals in rescuing configurations is significantly less acceptable when this anti-licenser is sentential negation. We do not know why this should be so, but it means that the best chance we could give to our alleged epistemic PPI modals would be to test for narrow scope under a subject NegDP—precisely the configuration banned by the Epistemic Containment Principle. Interestingly, Epistemic Containment does not prevent epistemic must from taking scope under (at least the negative portion of) NegDP subjects in contrastive/metalinguistic contexts. ii. A: Someone must have stolen my bird! B: No one must have stolen your bird; it’s entirely possible that it flew away on its own! iii. No one MAY have cheated. The numbers don’t lie; everyone MUST have cheated! More work is needed on this point.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



    

The observed wide scope of these modals cannot be achieved by lowering the anti-licenser, because negation does not move at LF (see Horn ; Penka and von Stechow ; Zeijlstra a; Abels and Martí ; Penka );¹³ nor does the negative portion of a NegDP (Iatridou and Sichel ). The  >  scope in (–) therefore cannot be achieved by lowering sentential negation or reconstructing the entire NegDP subject to its vP-internal position. Instead, Iatridou and Zeijlstra (, ) and Homer (, ) argue that the modal must move covertly to a position above the anti-licenser. This is head movement with an interpretive effect, independently argued for by Lechner  (see also Matushansky , especially Appendix ). Two characterizations of this movement have been proposed. Iatridou and Zeijlstra (, ) and Homer () describe it as a form of QR. As modals are quantifiers over possible worlds, it is unsurprising that they should be able to participate in QR just like phrasal quantifiers do. More particularly, Iatridou and Zeijlstra () propose that this operation moves an expression of type —the type of a modal quantifier whose restrictor has been saturated by a modal base—and leaves a trace of type s, the lowest type that will allow the structure to compose. This is shown schematically in the LFs for () and (), given in () and (), respectively. ()

LF: [PPI modalst,t  [NegDP [t2;s  [t1;s [vPs,t]]]]]

()

LF: [PPI modalst,t  [negt,t [t1;s [vPs,t]]]]

This is exactly parallel to what happens in phrasal QR, which moves an expression of type and leaves a trace of type e. Furthermore, like QR, this modal movement is clausebound (Homer ). ()

a. A professor greeted every student. b. A professor said that Sam greeted every student.

()

a. No one must leave. b. No one said that Chris must leave.

 > ,  >   > , * >  * > ,  >   > ,14 * > 

It is well known that certain DP PPIs are able to escape anti-licensing when they are spelled out beneath an anti-licenser, as demonstrated by the acceptability of () on the inverse scope reading; this is usually assumed to be achieved via phrasal QR. On this view, the mechanism that allows modal and DP PPIs to covertly escape anti-licensers is one and the same. ()

I didn’t understand something in your handout.

* > ,  > 

¹³ In any case, as Matushansky () observes, the semantic type of negation () makes it impossible for covert movement of negation to yield a semantic effect. ¹⁴ The  >  reading is acceptable in this case because the negative material is in a different clause from the PPI modal.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



Homer () argues that covert modal movement is crucially different from QR in that it is only available as a last resort, because modals do not take inverse scope when they are not under threat of anti-licensing. Homer demonstrates this for two environments. First, he shows that PPI modals cannot move over non-anti-licensers (). The string in (a) only has the surface scope reading, which does not accurately describe the instructions that John has been given; it cannot receive a felicitous inverse scope reading (cf. b). ()

John is an inexperienced cook, and to help him, Bill gave him very strict instructions about the dinner he is cooking for his girlfriend . . . a. #John often must stir this pot, otherwise the risotto will scorch. * >   >  b. John must often stir this pot, otherwise the risotto will scorch. (Homer : –)

Second, he shows that modals cannot move covertly over anti-licensers when they are in a shielding configuration. We have already seen in () that PPI modals are able to take scope under anti-licensers in this configuration, provided that an appropriate operator intervenes between them at LF. What Homer () demonstrates is that this is the only reading that is available, as shown in (). ()

Not everyone must jog.

 >  > , * >  >  (Homer : )

Homer therefore proposes that PPI modals move via a special last-resort movement operation, distinct from QR, that applies to prevent polarity items from being anti-licensed; he dubs this movement escape. It is not clear whether escape is intended to differ from QR aside from its last-resort status.¹⁵ Francis () notes that, while both of Homer’s cases show an unexpected unavailability of covert modal movement, they do not straightforwardly establish that this is due to modals moving by a special last-resort covert movement operation distinct from QR. First, if escape only forbids modals from moving when they are not in danger of being anti-licensed, we should expect covert movement to be available in cases parallel to () where the adverb is an anti-licenser. However, as () shows, this is not the case; wide scope is not available for must over anti-licensers like never any more than it is over non-anti-licensers like often. ()

*You never must give up.

* > , * > 

¹⁵ Homer () raises one other possible difference: covert modal movement violates Mayr and Spector’s (, ) Generalized Scope Economy, which bans covert scope-shifting operations from creating a stronger meaning than the surface scope interpretation. However, the surface scope structure in the cases under discussion contains an unlicensed PPI. Mayr and Spector (: fn. ) suggest that the surface structure is not a viable competitor to the structure that has undergone the covert scope-shifting operation when the former is not interpretable; indeed, they note that Fox’s () Scope Economy constraint, on which theirs is based, is formulated so that it only decides between two structures when both converge.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



    

Similarly, the absence of covert modal movement in the shielding configuration in () only constitutes evidence that the operation that covertly moves modals is different from QR if items that are known to move by QR are able to move over not every. However, () suggests that the latter is not the case: ()

Not everyone read three books.

 >  > , * >  > 

Finally, it should be noted that the fact that covert modal movement is only available as a last resort is only an argument against this movement being QR if there are no other restrictions on the position of modals at LF. We have already seen one such restriction that would derive the last-resort flavour of covert modal movement: Iatridou and Zeijlstra’s () claim that moved modals obligatorily reconstruct unless doing so would result in a polarity item being anti-licensed. They frame this as a condition on covert head movement, but acknowledge that the reasons for its existence are mysterious. In any case, both accounts agree that PPI modals are (only) able to move covertly in order to escape environments where they are not licensed. Homer’s arguments for escape center on environments where covert modal movement appears to be more restricted than the more familiar QR of DPs. It is worth noting that the availability of covert movement also differs from the availability of QR for DPs in the other direction. Francis (, ) demonstrates this point for sentences that have undergone negative inversion, a form of subject-auxiliary inversion accompanied by preposing of a negative expression illustrated in (). ()

Under no circumstances would I ever miss a phonology talk.

It has been claimed that the preposed expression in negative inversion constructions must take scope over everything else in the clause (Haegeman ; Büring ; Collins and Postal ; see also Potsdam ). That is, quantificational subjects cannot outscope the preposed expression—a surprising restriction on the covert movement of otherwise QRable expressions. This is true for both polarity-neutral and PPI DPs, as shown in (). However, PPI modals are able to outscope the preposed expression, as shown in (). ()

()

a. Never have more than four students passed this exam.

 >   , *   > 

b. *Never has someone in my department seen the ocean.

* > , * >  (Francis : )

To no student must you give the answers.

* > ,  >  (Francis : )

Francis () argues that this can be explained without assuming that the mechanism for covert modal movement is fundamentally different from QR if we make the following two assumptions: (i) QR obeys Superiority, as argued by Bruening (), and (ii) Relativized Minimality is sensitive to the difference between heads and phrases (Rizzi  et seq.). If QR obeys Superiority, the quantificational material in the preposed constituent will block

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



QR of the subject DP across it. However, if this locality condition is relativized to consider only material of the same phrasal type, we should not expect covert modal movement, which moves a quantificational head, to be blocked by the phrasal quantificational material in the left periphery.

... The source of polarity sensitivity The polarity theories delve a bit deeper than the positional theories of section ., in that they reduce the behavior of modals to that of the known phenomenon of polarity sensitivity. But they can only be truly explanatory if we understand what is behind the phenomenon of polarity sensitivity itself. This is a tall order: it involves answering many questions, such as why polarity sensitivity exists, what (anti-)licenses polarity-sensitive items, why modals should have polarity-sensitive behavior, and how polarity sensitivity in the domain of modals does or does not differ from polarity sensitivity in the domain of individuals.¹⁶ A discussion of polarity sensitivity is well beyond the scope of the current chapter; the literature on this topic is huge. Researchers have addressed both the question of what the licensing conditions of polarity items are and the question of what lexical properties are at the root of polarity itemhood. A longer version of the current chapter would examine accounts of polarity like Israel (), Giannakidou (), Chierchia (, ) (and many others), and explore whether and how they carry over to modal elements. Unfortunately, we cannot do justice to this topic in the current context, so we close this section with a few remarks only. There have been many attempts to explain what underlies the phenomenon of polarity sensitivity. For example, Szabolcsi () attributes NPIhood and PPIhood to the presence, number, and active/dormant status of [neg] features. This approach misses the oft-repeated generalization that polarity items are linked to scalarity, a link that plays a central role in the accounts of Israel (), Chierchia (, ), and Nicolae (). Israel () relies on lexically specified quantitative and informative properties of polarity items within a scalar model, while Chierchia (, ) and Nicolae () rely on the idea that polarity items obligatorily activate alternatives that must be exhaustified; the type of alternative and the method of exhaustification yield the difference between NPIs and PPIs. Any of these theories could be extended to modal polarity items. There is also a modal-specific account of the source of (positive) polarity sensitivity: Giannakidou and Mari () attribute the PPI status of epistemic must and its cross-linguistic counterparts to a lexical specification for a non-empty ordering source.¹⁷ On this view, the source of polarity sensitivity for modals is different from the source of polarity sensitivity for ¹⁶ For example, Iatridou and Zeijlstra () note that among modals one can find universal NPIs, unlike what is usually assumed in the domain of individuals. However, Zeijlstra (b) argues that one can find universal NPIs in the domain of individuals as well. ¹⁷ On Giannakidou and Mari’s view of modals, the ordering source for epistemic must requires a homogeneous set of worlds to be quantified over, and negation interrupts this homogeneity. More specifically, they propose that epistemic modal bases come partitioned into stereotypically ideal and non-ideal worlds. The modal quantifies over the ideal worlds. The ordering source is supplied by a (possibly covert) adverb merged above the modal structure; it ranks the ideal worlds higher than the non-ideal worlds with respect to speaker confidence (i.e. it states that the speaker is more confident in the ideal/stereotypical worlds than in the non-ideal/non-stereotypical worlds).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



    

other polarity-sensitive expressions. This is in contrast to the other accounts of polarity sensitivity mentioned above, which allow for a unified view of polarity sensitivity in modal and non-modal domains. It should be noted that Giannakidou and Mari’s () account says nothing about the distinction between polarity-neutral and NPI modals. Its coverage is thus narrower than the other accounts of polarity sensitivity just mentioned.

.. N-R

.................................................................................................................................. The final proposal that we will discuss seeks to link the wide-scoping properties of certain modals with the wide-scoping properties of another class of sentence-embedding predicates by attributing them to the same source: Neg-Raising. “Neg-Raising” refers to a phenomenon where certain predicates (e.g. think, want) receive a wide-scope interpretation with respect to a negation that appears above them in the surface structure, as demonstrated in () and (). ()

a. Dorabella doesn’t want to play hockey. b. Paraphrasable as: Dorabella wants to not play hockey.

()

a. Priscilla doesn’t think that it’s raining. b. Paraphrasable as: Priscilla thinks that it’s not raining.

At least two kinds of mechanisms have been proposed to account for Neg-Raising itself. The first one is also the one that gives the phenomenon its name, namely syntactic raising of negation out of the embedded clause (see Collins and Postal  and references therein). The second assumes that the negation originates in the matrix clause and derives the nonsurface scope interpretation via a homogeneity (also known as excluded middle) presupposition (Bartsch ; Horn ; Gajewski , i.a.). In a rough outline, this would work as follows. We presuppose that attitude holders are opinionated—that is, that attitude holders think either p or its negation, or want either p or its negation. Then, when it is asserted that it is not the case that the attitude holder thinks or wants p, we conclude that the individual in question thinks or wants the negation of p: ()

a. Presupposition: Priscilla thinks that it is either raining or not raining. b. Assertion: It is not the case that Priscilla thinks that it is raining. c. Enriched meaning: Priscilla thinks that it is not raining.

While in principle either of the above two mechanisms for deriving the Neg-Raising result could have been relevant, Homer () examines the Neg-Raising properties of modals from the homogeneity presupposition perspective. He argues that indeed certain modals scope over negation because they are PPIs, but others scope over negation as the result of Neg-Raising. More specifically, he argues that must is a PPI, but should is a Neg-Raiser. He offers two tests to distinguish PPI hood behavior from Neg-Raising behavior. The first is the availability of cyclic Neg-Raising. When a Neg-Raising predicate is embedded under a

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



doxastic Neg-Raiser like think, a negation in the matrix clause can be interpreted in the most embedded clause (Gajewski ). In (), we see that deontic should patterns like known Neg-Raiser want in () in allowing cyclic Neg-Raising, whereas must in () does not. ()

a. The doctor doesn’t think that John wants to jog. b. Paraphrasable as: The doctor thinks that John wants not to jog.

()

a. The doctor doesn’t think that John should jog. b. Paraphrasable as: The doctor thinks that John ought not to jog. (adapted from Homer : )

()

a. The doctor doesn’t think that John must jog. b. Not paraphrasable as: The doctor thinks that John is required not to jog. (Homer : )

(Homer : )

The second test that Homer () proposes to distinguish a Neg-Raising PPI from a nonNeg-Raising PPI is the availability of a wide scope existential quantification reading for not everyone. This reading, illustrated in () with the known Neg-Raiser want, is a hallmark of Neg-Raising constructions. Homer claims that this reading is available for should but not for must, as demonstrated in () and (), respectively.¹⁸ ()

a. Not everyone wants to help me. b. Paraphrasable as: There is some x such that x wants not to help me. (Homer : )

()

a. Not everyone shoulddeo get a flu shot. b. Paraphrasable as: There is some x such that x should not get a flu shot. (Homer : –)

()

a. Not everyone mustdeo get a flu shot. b. Not paraphrasable as: There is some x such that it is required that x does not get a flu shot. (Homer : )

Thus, according to Homer (), covert movement is not the only way in which modals can achieve wide scope over negative material; at least some modals scope over negation by the mechanisms underlying Neg-Raising as well. The Neg-Raising approach, like the polarity-sensitivity approach discussed in section ., seeks to reduce the wide-scoping properties of certain modals to a phenomenon that exists elsewhere in the grammar. Although Neg-Raising does not explain why some modals obligatorily scope under, or require the presence of, negation, it does provide a way of explaining subtle differences between modals that scope over negation that polarity

¹⁸ A reviewer reports that the facts in British English may be different on this point.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



    

sensitivity alone cannot address. Taken together, Neg-Raising and polarity sensitivity deliver an intricate pattern of scopal interactions between modals and negation.

.. C

.................................................................................................................................. In this short chapter we have reviewed three ways in which researchers have proposed to explain the patterns that result from the scopal interaction of negation with modals: (i) different positions of interpretation, (ii) polarity sensitivity, and (iii) Neg-Raising. We hope to have clarified certain strengths and weaknesses of each approach and to have shown that while our understanding of these issues has increased over the years, many challenges remain, not the least of which is a broader cross-linguistic investigation of this interaction.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  ......................................................................................................................

    ......................................................................................................................

 

.. ‘N’      

.................................................................................................................................. I a world of events as dense as the spatiotemporal regions they occupy, much of what might be said to describe them in a language with negation threatens to be trivial (Partee ; Burge ; Ogihara ). During the flight, it was bumpy, not calm ().¹ False then that it was calm and not bumpy (). Yet, even the worst flight has its calm spots, although these are too scattered to falsify () or to verify (): ()

It was not calm.

()

F It was not bumpy.

The logical form of negation in () must not be (), if a spot of calm among the events E of the flight is not to make it trivially true. Rather, () says that not any of the flight events E are bumpy, as in ():²

¹ In ()–(), an impersonal construction, with weather-‘it’, to postpone the confound of additional arguments. ² Monadic second-order logic (v. Schein ), with restricted quantification. First-order, lowercase variables in argument position, and second-order, uppercase variables in predicate position: (i) Aristotle is some thing, e.g. a mortal. (ii) Aristotle is something, e.g. mortal. 9X Xa, e.g. [℩X: 8x(Xx $ Mx)]Xa 9x(a = x), e.g. [9x: Mx](a = x) Notational conventions: i. Concatenation without parentheses for predication of second-order variables or constants, ‘Xx’, ‘Ee’, ‘Ma’; ii. Lexical predicates and relations occur with variables in parentheses, ‘mortal(x)’, ‘calm(e)’, ‘Agent(e,x)’, ‘Cause(e,e’)’; iii. Brackets append to a molecular formula ‘Φ[ . . . ]’ to indicate that the enclosed variables occur free within it; iv. Spoken all, some, a(n), the, etc. appear as such in logical form, reserving ‘8’, ‘9’, and ‘℩’ for inaudible, interpolated logical form.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

()

* 9e ¬ bumpy(e); *[9e: Ee] ¬ bumpy(e)

()

¬ [9e: Ee] bumpy(e)

If, again however, the flight itself were events E as dense as the spatiotemporal regions it transits, a scattered calm spot would falsify () and its logical form (), which would be mistaken for a report of extreme weather: ()

¬ [9e: Ee] calm(e)

Were the flight through E dense with events e, each an arbitrarily small slice of air space and flight time and none calm, the aggregate turbulence would surely have downed it (Ogihara ).³ For () true, it must rather be that air space and flight time are parsed into events E worthy of notice for air traffic, and each e among these E is a flight segment too large not to have suffered at least a few bumps. Reference to the events E is often implicit and demonstrative for () verbatim (Partee ; Burge ).⁴ Yet, a spatiotemporal frame adverbial is explicit description of framing E as in (), and the relation between the framing E and the framed e is an asymmetric one, () vs.(), a framing richly attested in adverbial modification elsewhere as in () (Taylor ; Davies ; Schein ): ()

Once upon an unknown dark and stormy night, it was not calm. [9E: Once upon a dark and stormy night[E]] ¬[9e: Ee] calm[e]

()

In flight for two hours,⁵ it wasn’t calm for two minutes. [9E: in flight for two hours[E]] ¬[9e: Ee] calm for two minutes[e]

Pluralization. For first-order event concepts V, first-order temporal, causal, or mereological relations R between events, and first-order thematic relations θ (v. Schein , ): ⌜V[E]⌝ for ⌜9eEe & 8e(Ee ! Ve)⌝ ⌜R[Ei,Ej]⌝ for ⌜9eEie & 9eEje & 8ei(Eiei ! 9ej(Ejej & Reiej)) & 8ej(Ejej ! 9ei(Eiei & Reiej))⌝ ⌜θ[E,X]⌝ for ⌜9xXx & 9eEe & 8x(Xx $ 9e(Ee & θex))⌝ (Pietroski ) E.g. ⌜calm[E]⌝ for ⌜9eEe & 8e(Ee ! calm(e))⌝. Given ⌜Agent(e,x)⌝ to mean ‘x is an Agent of e’, ⌜Agent[E,X]⌝ means ‘X are the Agents of E’. For any predicate V or relation R, ⌜V[vi]⌝ for ⌜9Vi(8v(Viv $ v = vi) & V[Vi])⌝ ⌜R[vi,Vj]⌝ for ⌜9Vi(8v(Viv $ v = vi) & R[Vi,Vj])⌝ ⌜R[Vi,vj]⌝ for ⌜9Vj(8v(Vjv $ v = vj) & R[Vi,Vj])⌝ ⌜R[vi,vj]⌝ for ⌜9Vi9Vj(8v(Viv $ v = vi) & 8v(Vjv $ v = vj) & R[Vi,Vj])⌝ E.g. ⌜Agent[e,X]⌝, ‘X are the Agents of e’; ⌜Agent[E, x]⌝, ‘x is the Agent of E’; ⌜Agent[e, x]⌝, ‘x is the Agent of e’. Note that for a primitive, monadic predicate ⌜Vvi⌝ and ⌜V[vi]⌝ are equivalent, and so, both ‘calm(e)’ and ‘calm[e]’ passim. ³ Storm chasers fly through, in, and around storms. A flight is through some events just in case the flight is through the spatiotemporal regions the events occupy. ⁴ Burge (:  n. ) cites Russell (: ch. ) and Reichenbach (: –) for the insight that tensed constructions contain a demonstrative element. ⁵ In flight for two hours represents an open class of spatiotemporal frame adverbials including: aloft dawn to dusk, during that -hr flight, across Southern New England, flying yesterday, being stuck in first class the entire way, when I was the pilot, while I was drinking martinis nonstop for two hours on a two-hour non-stop flight, etc.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   



() # In flight for two minutes, it wasn’t calm for two hours. [9E: in flight for two minutes[E]] ¬[9e: Ee] calm for two hours[e] ()

In a perfect harmony that shook the cathedral, each of a thousand voices breathed a solitary, mournful note of a minor, broken chord.

The logical form thus considered for () divides it as in () between a description Φ of the framing events E and a point-wise description Ψ of the framed events e: () In flight for two hours, it was not calm. [9E: in flight for two hours[E]] ¬[9e: Ee] calm[e] () [9E: Φ[E]] ¬ [9e: Ee] Ψ[e] But, with no reading of () equivalent to (), the first-order quantification over singular e apparent in () and withheld from () must owe its presence to ‘not’ and not to anything () and () hold in common: ()

In flight for two hours, it was calm. [9E: in flight for two hours[E]] calm[E]

()

In flight for two hours, it was sometimes calm. [9E: in flight for two hours[E]] [9e: Ee] calm[e]

That is, ‘not’ itself is the event quantifier, with () equivalent to (), a thinly disguised adverb of quantification—noughtly says it like it is—conforming in its logical syntax and semantics to that of other adverbs—mostly, rarely, always, never, sometimes, often⁶:

⁶ Noughtly is most like mostly, congenial to mass and count—most time(s), most space(s)—and unbiased toward the temporal or spatial. In both respects, noughtly and mostly differ from never and usually, as an anonymous reviewer points out, in that the latter are biased toward a count domain of temporally individuated events. Compare: (i)

a. Agent ’s accent is mostly Russian. b. Agent ’s accent is usually Russian.

(ii) a. Agent ’s accent is not Russian. b. Agent ’s accent is never Russian. Despite their differences, it is uncontroversial that mostly and usually are adverbs of quantification and both say about their respective domains in (i) “more Russian than not”—much as each, every, any, and all are restricted, universal quantifiers without synonymy (Vendler a). The anonymous reviewer notes further important contrasts between not and never: (iii) (iv) (v) (vi)

In flight for two hours, it wasn’t calm. In flight for two hours, it was never calm. Agent  is not at home. Agent  is never at home.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

()

[9E: in flight for two hours[E]] [No e: Ee] calm[e]

()

In flight for two hours, it was mostly calm. [9E: in flight for two hours[E]] [Most e: Ee] calm[e]⁷

These quantifiers are restricted as in ()–() to a domain of events decided in a context that decides on a spatiotemporal region and its event segmentation into events E:⁸ ()

(Negation in Event Semantics) [9E: Φ[E]] [No e: Ee] Ψ[e] [℩E: Φ[E]] [No e: Ee] Ψ[e]8

.. A     

.................................................................................................................................. A few observations have fixed () as the canonical logical form of negation. In event semantics, ‘not’ is not a sentential connective. It belongs to a different part of speech, adverbs that include mostly, rarely, etc., restricted event quantifiers. The part of speech dictates in turn the surrounding syntax in (). As an adverb of quantification, there must be reference to its restriction, framing events E that Φ[E], whether in spoken description as in () or as a silent demonstrative in (). As a corollary, the logician’s ⌜¬Ψ⌝ is no sentence of natural language, not even (). Negation is never leftmost in a complete thought. To understand a sentence parsed as in (), when the events E of Φ-ing are plainly not intended to be as dense as the spatiotemporal regions they occupy, it must be decided what event segmentation is intended. For negation in event semantics, there is thus far: i. its part of speech as the adverb noughtly, ii. a canonical logical form (), and iii. some regime for event segmentation. Nothing yet regiments clauses ⌜Φ[E]⌝ and ⌜Ψ[e]⌝ in (). Doing so revisits the foundations for the semantics and logic of negation in natural language, especially in talk about events. Negative descriptions of events, ⌜not Φ[E]⌝ and ⌜not Ψ[e]⌝, face problems both logical (Perceptual Veridicality and Consistency) and semantical, the problem of Negative Events. Consider the storm that did as () says: Sentence (iv) reports a bumpier two hours than (iii) does; and (vi), in contrast to (v), cannot mean that no state of Agent  now is her being at home now. Both contrasts are expected if never derives from negation plus negative polarity minimizer, i.e. never < notþever = ‘not any a moment at all (ever anywhere)’. Noughtly is the purest event quantification, presupposing neither count nor mass, neither spatial nor temporal, neither fine nor gross segmentation, neither realis nor irrealis. Classifying negation as either sentential connective or adverb of quantification does not explain its unique syntax in association with verbs, e.g. morphological affixation, negative concord, etc., as Terje Lohndal has reminded me. That is, logical syntax lags behind natural language, as elsewhere: the classification of nouns, adjectives, and verbs as monadic predicates also fails their syntactic distribution. ⁷ In ()–(), mostly and noughtly are translated as nominal event quantifiers ‘[Most e: Ee]’ and ‘[No e: Ee]’, although I would rather it be ‘[Mostly e: Ee]’ and ‘[Noughtly e: Ee]’ to anticipate rarely, often, usually, which are without ready nominal counterparts. ⁸ Spatiotemporal frame adverbials occur as both indefinites, once upon a dark and stormy night, and definites, that dark and stormy night.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   



()

The storm tripped the “Fasten Seat Belt” light and iced the wings, and the passengers felt the storm.

()

It not being calm tripped the “Fasten Seat Belt” light.

()

It not being dry iced the wings.

()

The passengers felt it not being calm.

()

F It not being dry tripped the “Fasten Seat Belt” light.

()

F It not being calm iced the wings.

()

F The passengers felt it not being dry.

Since the storm s did all these things, the negative event descriptions in ()–() do not describe it, and the ‘not’ they contain violates Excluded Middle: ()

¬ being dry[s] & ¬ not being dry[s]

()

¬ being calm[s] & ¬ not being calm[s]

Not ‘¬’, this ‘not’ has the power to describe a negative event that is no storm and yet powered enough to trip the light, ice the wings, or unsettle the passengers, ()–(). With () and () equivalent, it looks equally beyond noughtly to empower negative events, and a special meaning for the ‘not’ in these contexts is alleged (section .). Defying Excluded Middle, negative event descriptions trouble logic further. Perceptual Veridicality subsumes inference from feeling it not be calm (v. ()) to it not being calm simpliciter. Yet, perception of an event of it not being calm or even of the event of it not being calm should not inhibit the presence and perception of a different event of it being calm, denying the purported inference. According to Perceptual Consistency, if one feels it not be calm, then one does not feel it be calm. But, again, why should feeling the negative event preclude feeling a positive one too? Whether ‘not’ is noughtly or ‘¬’, there is not yet explanation for the violation of Excluded Middle, Negative Events, or the elusive Perceptual Veridicality and Consistency. Nor, given the equivalence via () and () to classical negation, does the adverb of quantification readily explain violations of Excluded Middle that appear to appear without reference to events and to engage instead plural or mass reference. Of scattered rains, some warm and some not, the rains or rain are thought neither warm nor not warm (¬ warm[R] & ¬ not warm[R]), where again a special condition is alleged to stipulate when negation violates Excluded Middle (section ..). Addressing these foundational problems, a logical language for event semantics takes on three more features besides noughtly and event segmentation. First (section .), if () is canonical, then () is what is embedded, ⌜[the E : ()]⌝, in plural event definite descriptions— ‘the time/events in flight for two hours, that it wasn’t calm’, ‘when/where in flight for two hours, it wasn’t calm’. Second (section .), logical form conforms to the neo-Davidsonian thesis of

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

thematic separation, viz., that all nominal arguments are arguments in their own, separate relation to events, spoken, as in close to the clouds ‘[the X: clouds(X)]close-to[E,X]’, or left silent as in ‘[the X: clouds(X)] Theme[E,X]’. Neo-Davidsonian logical form enlarges the inventory of phrases that occur as spatiotemporal frame adverbials, Φ in (), (v. section ..); and, in separating all nominal arguments from their verbal or adjectival predicates, it translates the latter as monadic event concepts (v. section ..). Third (section .), within the scope of any adverb of quantification quantifying over events e as in (), definite descriptions (and pronouns as definite descriptions) are restricted to those described that are local to the event e (localized definite description). So parsed, the language of event semantics resolves the foundational problems. Except for the finding that ‘not’ is an adverb of quantification, there is no grammar or theory special to negation, with the three additional features just listed justified independently. Negation is unambiguous in all its tokens, its logic is classical, and there is not a plenitude of negative events. What is interesting about negation in event semantics are the reductive and eliminative consequences of its translation into event semantics.

.. N-D 

..................................................................................................................................

... Thematic separation Canonical logical form () distinguishes the framed e from the framing E and leaves it to grammar which of a sentence’s phrases may instantiate Φ or Ψ, so that ()–() and () turn out unambiguous but () ambiguous in its description of either the framed e or the framing E: ()

For two hours in flight, it wasn’t calm.

()

It wasn’t calm for two hours in flight.

A neo-Davidsonian (Castañeda ; Parsons ; Schein ; Lohndal ; Williams ) language for event semantics enlarges the inventory of phrases that grammar may call upon for Φ and Ψ: ()

Close by the clouds for two hours, it wasn’t calm for two minutes (underneath).

()

[The X: clouds(X)][9E: close-by h[E,X]] [No e: Ee] calm m[e]

()

The clouds close by for two hours weren’t calm for two minutes (underneath).

()

[The X: clouds(X) & [9E: close-by h[E,X]][9E: Theme[E,X]] [No e: Ee] calmm [e]

The apparent subject predication in () is a frame adverbial in light disguise in which a thematic relation, ‘Theme’ in (), replaces ‘çlose-to-for--hours’ in (). There is then little to tell apart the logical forms for () and () except for the noted substitution (v. Schein  for further discussion).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   



... Monadic event concepts Neo-Davidsonian logical form matters twice. First, as above, it assimilates ‘[the X: clouds[X]] Theme[E,X]’ and any other prepositional phrase. Second, translating a predicate as a monadic concept of events, ‘calm(e)’, dissociates what it denotes from what participates therein and how. Even when the wall tiles are each flat (), their misalignment at a seam results in a state that is not, (). That is, the tiles’ flatnesses are not the wall’s. Flat themselves, something else isn’t flat, to which the tiles stand in some relation other than identity. ()

The tiles are not flat.

()

8x(tile(x) ! flat(x))

()

9e ¬flat(e)

Abbreviating contextual restriction to the wall’s state as ‘wall-sized[E]’⁹, the logical form for () denies its flatness alone, leaving the tiles intact: ()

[The X: tiles[X]][9E: [E] & Theme[E,X] & wall-sized[E]] [No e: Ee] flat[e]

With such truth conditions for (), a monadic event concept dissociated from the event’s participants is essential for the scope of negation. Given that a univocal concept flat subsumes all that it fits, when the tiles are each flat, () must look elsewhere, to the state of the wall, for that which isn’t flat, as in ().¹⁰ In accepting (), rejecting (), one understands something about a first-order, singular concept, ‘flat(v)’, viz., () that a flatness is a geometry implying the flatness of its constituents (whereas the geometry of a swirl does not imply swirls within):¹¹

⁹ The wall-sized states E in this context are just the one of the wall, and I allow with ‘[E]’ that just one may be intended: [E] $ def 9e8e’(Ee’ $ e’ = e). ¹⁰ More fun, (i) with one concept swirl describes hurricanes each a swirl in a stationary cell that is not a swirl of swirls, and (ii) shows that the monadic predicate expresses a concept of events that the adverb swiftly may in turn modify: (i) The swirls do not swirl. (ii) The swift swirls do not swirl swiftly. (iii) [The X: swift[X] & swirls[X]][9E: [E] & Theme[E,X] & cell-sized[E]] [No e: Ee] (swirl[e] & swift[e]) ¹¹ The primitive, monadic concept of events may take on board arbitrary distributional or topological properties defined over the event’s participants: (H) (From Healthy Schools Rules & Regulations) Theme[e,X] & healthy(e) & Nanuet School District[e] ! . . . The health records of its pupils X determine the health of a school population, which is healthy, ‘healthy (e)’, according to standards setting maximum weekly, daily, classroom, and school-wide absentee rates that vary for elementary and secondary schools and set further epidemiological and environmental standards. Unaware of the details, it may nevertheless be reported that not a school is healthy today:

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

()

The tiles are a wall. The wall is flat. Therefore, the tiles are each flat.

()

The storms are a storm cell. The storm cell swirls. #Therefore, the storms each swirl.

()

(Flatness) (Theme[e,X] & flat(e)) ! [8x: Xx] 9e(Theme[e,x] & flat(e))

Suppose the flat tiles T are not flat on a crooked wall w’, which is then repaired so that the same tiles T are flat on a flat wall w: ()

The tiles are not flat. [The X: tiles[X]][9E: Theme[E,X] & wall-sized[E]] [No e: Ee] flat[e]

()

The tiles are flat. [The X: tiles[X]][9E: Theme[E,X] & wall-sized[E]] flat[E]

Given that [8t: Tt]flat(t), some might cast () as ‘¬flat(T)’, appealing to a collective, plural predicate (v. section ..), or better as ‘¬flatw’(T)’ so as not to contradict () as ‘flatw(T)’. Even so, there is no collective, plural predicate flat in (). If it is the same concept flat throughout, it must be as in (), in accord with the event semantics for collective predication: ()

flatw(T) $def Theme[w,T] & flat(w)

That is, the tiles surface w (Theme[w,T]), the wall, a spatiotemporal region, event, or state that conforms to the geometry of a flatness, and thus subject to ().

.. L  

.................................................................................................................................. Description of the framing E may include landmarks or waypoints for E such as the clouds scattered across Southern New England along a two-hour flightpath: ()

Close to the clouds of Southern New England for two hours, it wasn’t calm underneath (them) for  minutes.

(i) The pupils of the Nanuet School District are not healthy today. (ii) [The X: pupils[X]][9E: Theme[E,X] & school-sized[E]] [No e: Ee] (healthy[e] & today[e]) On any day at any one school, there is exactly one event at issue, the health of its population. It is not a state of health, deviating from (H). And, (iii) reports states of health equinumerous with the district’s schools, without implying that the reporter or context for the report has access to their number, the partition of pupils among them, or to much about (H): (iii) The pupils of the Nanuet School District were healthy yesterday. (iv) [The X: pupils[X]][9E: Theme[E,X] & school-sized[E]] (healthy[E] & yesterday[E])

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   



()

Aloft for  hours, the clouds of Southern New England weren’t calm underneath (them) for two minutes.

()

Aloft for  hours, it wasn’t calm underneath the clouds of Southern New England for two minutes.

Description of the framing E for the flight joins a point-wise description of the local e for the flight segments en route, for which the clouds are waypoints along the way. There is no thought that reference en route to the clouds squeezes Southern New England’s clouds into a -minute flight segment. That is, the definite description in () and the pronouns in ()–() spoken and unspoken are all understood as localized definite descriptions restricted to the local e, in which respect noughtly is like other adverbs of quantification and modal operators capturing the definite descriptions within their scope: ()

. . . [9E: . . . for  hr[E]] [No e: Ee] . . . [The X: SNE clouds[X] & local[e,x]] . . . [e] . . . the clouds of Southern New England that are at e . . .

Localized definite description is also plain in unnegated counterparts to ()–(), which may be true despite the constant turbulence roiling Southern New England clouds all day if the narrow flightpath escapes it en route: ()

Close to the clouds of Southern New England for two hours, it was calm underneath (them) (for  minutes at a time).

()

Aloft for  hours, the clouds of Southern New England were calm underneath (them) (for two minutes at a time).

()

Aloft for  hours, it was calm underneath the clouds of Southern New England (for two minutes at a time).

()

[9E: . . . for  hr[E]] ( . . . [8e: Ee]calm[e] & [8e: Ee][The X: SNE clouds[X] & local[e,x]]underneath[e, X] . . . ) . . . the clouds of Southern New England en route at e . . .

The Southern New England clouds are never calm en masse. Yet, at every waypoint along the way, those then there were, as () says. The localizing effect of various distributive operators on the definite descriptions in their scope is fully general and independent of facts about negation (v. Schein  for discussion).¹² With ‘not’ as noughtly, event ¹² It cannot be overestimated how easily definite descriptions localize: (i) The medieval dioceses married the peasants in the local parish church. The dioceses, marriages, and peasants are scattered across the ages and Christendom without (i) siting them in the same one church local to them all. Rather, (ii) . . . married[E] . . . & [8e: Ee][The x: local[e,x] & church[x]] in[e,x] . . . “in the local parish church”

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

semantics moves only to append negation to the list of contexts that induce it, evidently so in ()–() and essential to dissolve an illusion from negation that definite description is ambiguous (section ..).

.. D     

.................................................................................................................................. The logical language to which () belongs includes plural descriptions. Event semantics has plural event descriptions, among which are negative descriptions of events. The definite descriptions ()–(), instances of ⌜[the X : ()]⌝ or ⌜[the E’ : ()]⌝, embed and abstract on the italicized instances of () shown above them, as expected in a language with plural definite description and clauses the shape of (): The clouds above Southern New England for  hours weren’t calm underneath for  minutes—

()

the clouds above Southern New England for  hours that weren’t calm underneath for  minutes.

()

a. the airspace(s) where/ flight time(s) when the clouds above Southern New England for  hours weren’t calm underneath for  minutes. b. where/when(ever) the clouds above Southern New England for  hours weren’t calm underneath for  minutes.

Close to the clouds of Southern New England for  hours, it wasn’t calm underneath for  minutes—

()

a. the airspace(s) where/ flight time(s) when close to the clouds of Southern New England for  hours, it wasn’t calm underneath for  minutes.

b. where/when(ever) close to the clouds of Southern New England for  hours, it wasn’t calm underneath for  minutes. Aloft for  hours, it wasn’t calm underneath the clouds of Southern New England for  minutes—

()

a. the airspace(s) where/ flight time(s) when aloft for  hours, it wasn’t calm underneath the clouds of Southern New England for  minutes.

b. where/when(ever) aloft for  hours, it wasn’t calm underneath the clouds of Southern New England for  minutes. The descriptions ()–() cull their reference from across Southern New England from the local frames of reference en route, referring to all and only the clouds () or all and only the airspace or flight time ()–() disturbed by turbulence—fewer than all the clouds of Southern New England and less than the airspace and total flight of two hours but more than the clouds or the airspace and flight time of a single two-minute segment that wasn’t a calm. The abbreviated logical form for () (v. Schein ) abstracts on ‘X’ and quantifies into a definite description relativized to the local event e, as the context for localized definite

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   



description requires. Logical forms for ()–() abstract on ‘E’ ‘, culling events from across Southern New England: ()

[The X : clouds[X] & [9E: above SNE for  hours[E,X] & Theme[E,X]] [No e: Ee](calm[e] & [℩Y: [8y: Yy]Xy & local[e,Y]] under[e,Y] & for  min[e])] ‘the clouds X among which there was not a moment of calm for  minutes underneath those Y of them-X in that moment.’

()

[The E’: [The X: clouds[X] & [9E: above SNE for  hours [E,X]][9E: Theme[E,X]] [No e: Ee & local[E,E’]](calm[e] & [℩Y: [8y: Yy]Xy & local[e,Y]] under[e,Y] & for  m[e])] ‘where/when E’ the clouds of Southern New England had not a moment of calm then-E’ underneath the clouds in that moment.’

()

[The E’: [The X: clouds[X]][9E: close-to[E,X] & for  hrs[E]] [No e: Ee & local[E,E’]](calm[e] & [℩Y: [8y: Yy]Xy & local[e,Y]] under[e,Y] & for  m[e])] ‘where/when E’ close to the clouds of Southern New England for  hours, there was not a moment of calm then-E’ underneath the clouds in that moment.’

()

[The E’ : [9E: aloft for  hrs[E]][No e: Ee & local[E,E’]] (calm[e] & [the Y: SNE clouds[Y] & [8y: Yy]Xy & local[e,Y]] under(e,Y) & for  m[e])] ‘where/when E’ aloft for  hours, there was not a moment of calm then-E’ underneath the clouds of Southern New England in that moment.’

Examples of plural definite description via such abstraction pervade, including descriptive anaphora (Schein : ff.; ff.; ff.; ff.). Suppose farmers feed donkeys onepound bags of oats as depicted in ().¹³ Verifying ()–(), anaphoric reference to the events described by the antecedent sentence and to the farmers, donkeys, and oats therein omits the last two, afternoon events in (): ()

d d f< >b— min (early am) f< >b — min (late am) d d f—d—b— min (early pm) f—d—b — min (late pm)

()

a. Exactly two farmers each fed two donkeys one bag of oats. The oats weighed two pounds, and the donkeys ate for twenty minutes. b. Exactly two farmers each fed two donkeys one bag of oats. They were two pounds of rolled grain, and they ate for twenty minutes.

()

a. Exactly two farmers each fed two donkeys one bag of oats, in  minutes. b. Exactly two farmers each fed two donkeys one bag of oats. It took  minutes.

¹³ The early am graph, e.g., depicts farmer f feeding donkeys, d and d, bag of oats b in exactly  min.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

The descriptive anaphora in ()–() cull their reference to feedings, donkeys, or bags of oats from across events of a farmer each giving two donkeys one bag of oats and from no other events: ()

[the Z: bags of oats[Z] & [Exactly x: farmer(x)] 9e(Agent(e,x) & feed(e) & [9Y:  donkeys[Y]]To(e,Y) & [℩W: [8z: Wz]Zz & local (e,W)] Theme(e,W))] ‘the bags of oats-Z that exactly two farmers (each) fed to two donkeys those of themZ at hand’

The anaphora in ()–() referring to oats refers to the two bags b and b. Similarly, in interpreting the other descriptive anaphors in ()–(), it affords reference to just donkeys d, d and d and what they do in just the twenty minutes of the two morning events. These definite descriptions are unremarkable in their logical syntax given their abstraction on () and the logical syntax of quantifying-in localized definite descriptions (section .), and their meaning is plain enough (and defined in Schein  §: –; ). They constrict reference to just those clouds, airspace(s), or flight time(s) that witness turbulence—the not being calm segments—or to just the farmers, donkeys, oats, or times of ()’s morning.

.. N

.................................................................................................................................. Mistaken for something other than noughtly, negation perplexes.

... The homogeneity condition Both () and () may report a flaw in the wall: ()

Across (the surface of) the tiles, it is not flat.

()

[The X: tiles[X]][9E: across[E,X] & wall-sized[E]][No e: Ee] flat[e]

()

The tiles are not flat.

()

[The X: tiles[X]][9E: Theme[E,X] & wall-sized[E]][No e: Ee] flat[e]

A single tile askew verifies the report. Absent a wall and the tiles scattered on the floor, event segmentation resorts to them to individuate the intended events. In such a context, any dimension across the tiles is a dimension across a single tile. In such a context, the relevant events are taken to be as many as the tiles themselves so that () and () understood as ()–() are true just in case none of the tiles is flat:¹⁴ ¹⁴ v. Schein  for formal definition of event segmentation by individuation in relation to objects. For the tiles X to induce it, the events E are equinumerous, i.e. E are one-to-one with X, and there is an automorphism h from X to E: 8x(x is within the spatiotemporal region that h(x) occupies & 8y(Xy ! (y

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

    ()



[The X: tiles[X]][9E: across[E,X] & tile-sized[E] & One-on-One[E,X]][No e: Ee] flat[e]

() [The X: tiles[X]][9E: Theme[E,X] & tile-sized[E] & One-on-One[E,X]][No e: Ee] flat[e] In that same context and event segmentation, () and (), without negation, say of equinumerous events that they are flatnesses all (v. (), n. ), which is true just in case every tile is: ()

Across (the surface of) the tiles, it is flat.

()

[The X: tiles[X]][9E: across[E,X] & tile-sized[E] & One-on-One[E,X]] flat[E]

()

The tiles are flat.

()

[The X: tiles[X]][9E: Theme[E,X] & tile-sized[E] & One-on-One[E,X]] flat[E]

In such a context, if some tiles are flat and some are not, neither the negated sentences () and () nor the unnegated () and () are true.¹⁵ Now if one ignores () and () with an overt spatiotemporal frame adverbial, suppresses the event semantics that unifies them with () and (), and mistakes the effects of event segmentation for direct predication of equinumerous objects, () and () are cast as () and () to create the illusion that Excluded Middle has been violated, given that neither () nor () is true: ()

¬Flat(T)

()

Flat(T)

Since Fodor ,¹⁶ negation has worn a Homogeneity Condition custom made to the effect that homogeneous predicates under negation (() without a wall, () below) denote homogeneously—all are not so—in contrast to non-homogeneous predicates under negation (() with a wall, () below), for which it suffices that merely some are not so: ()

The tiles are not flat on one side. ¬FlatSide(T)

is within the spatiotemporal region that h(x) occupies ! y = x))), i.e. E are one-on-one with X, One-onOne[E,X]. ¹⁵ Given bias that plural event quantification favors plural, count quantification (v. Schein : §...), an anonymous reviewer observes that mass reference to the tile undermines plural event segmentation, so that (i) and (ii) are about the wall if there is one and less transparently about the tiles separately if they are scattered on the floor: (i) Across (the surface of) the tile, it is flat. (ii) The tile is flat. ¹⁶ v. e.g. Breheny ; von Fintel ; Fodor ; Gajewski ; Higginbotham ; Krifka ; Križ ; Löbner ; ; Lønning ; Magri ; Malamud ; Roeper ; Schwarzschild ; Spector ; Szabolcsi and Haddican ; Yoon .

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

()

The tiles are flat on one side. FlatSide(T)

()

The tiles are not flat across the wall. ¬FlatWall(T)

()

The tiles are flat across the wall. FlatWall(T)

Accordingly, the negation of homogeneous predicates defies Excluded Middle (() & ()), but the negation of non-homogeneous predicates does not (() & ()). The Homogeneity Condition is disguised as grammar (and less an accident of event segmentation) in standard citations that contrast intrinsically homogeneous predicates—The tiles are ceramic—with collective predicates, The tiles touch. Overlooked in this opposition is an intermediate condition: ()

[The two walls your guy did—] The tiles are flat.

()

[The two walls your guy did—] The tiles are not flat.

Of tilework on two walls, the unnegated () requires, as above, that all tiles are flat and none askew from its wall. For negated (), the tiles need not all not be flat, the predicate not being homogeneous. Yet, a single tile askew is not enough for (), unlike when the nonhomogenous predicate in () featured one wall. Here, () needs two tiles askew, one on each wall. With only one askew on one wall, () and () are neither true and again violate Excluded Middle, eluding the Homogeneity Condition that expects violations of Excluded Middle only from homogeneous predicates, the fully distributive ‘¬FlatSide(T)’, and Excluded Middle from all others. In () and (), it is simply that the events E in play are two installations on two walls, not a one e of which, () says, is a flatness, demanding a tile askew on each. The alleged homogeneity of () and () is just the coincidence that the framing events E are coincident with the tiles, and, says (), not an event e among these E is flat(e), which cannot be said without a monadic concept of events to say it. Absent event semantics, the presumption that ‘¬Flat(T)’, ‘¬FlatSide(T)’, and ‘¬FlatWall(T)’ distill the structure of negation has been occasion for an escalating arms race (v. op. cit. n. )—Boolean algebra, lexical presuppositions, Strongest Meaning Hypothesis, trivalence, supervaluation, double strengthening, etc.—ingenious in its evasion of the syntax and semantics of natural language.

... Plural definite descriptions under negation Plural definite descriptions (v. op. cit. n. ) suffer the illusion of an ambiguity between a universal meaning that demands calm under every cloud and an existential one that emerges under negation to demand there be calm under not a one: ()

It was calm under the clouds of Southern New England.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

    ()



It wasn’t calm under the clouds of Southern New England.

The illusion derives from definite description restricted to the local event e as in (). Absent noughtly, an adverb of quantification quantifying over such local e, there is little except to entertain ambiguity in the and the grammar of negation elaborated so that one the occurs only within its scope (v. op. cit. n. ). An example discriminates between localized definite description and the alleged existential meaning. Suppose the clouds are scattered in tight clusters across Southern New England, transit time underneath any cluster is  minutes, with calm air on the port side and turbulent air on the starboard side. It is therefore not calm under any cluster, but calm under some of its clouds, those on the port side. This context falsifies the unnegated (), which demands calm all along the flight path, as expected from the definite description. And, this context verifies () and falsifies (), demonstrating that the definite description in () is restricted to the local two-minute transit, refers to all the clouds in the local cluster, and is not existential: ()

Aloft  hours, it was calm under the clouds of Southern New England.

()

Aloft  hours, it wasn’t calm under the clouds of Southern New England (for even  minutes).

()

Aloft  hours, it wasn’t calm under any (of the) clouds of Southern New England for (even  minutes).

That is, the is unambiguous throughout and its host definite description is restricted to the local event when it falls within the scope of event quantification.¹⁷,¹⁸

.. N     

.................................................................................................................................. If outside and in the seat of their pants denote events, () describes an event felt that is not the event of feeling it, the former being outside and the latter inside. The “obvious” (Vlach ) theory proposes () (Higginbotham : ; Higginbotham ; Parsons ; Vlach ), with an indefinite description of the object of perception, and in () of the instrument of causation:

¹⁷ Beyond negation proper, the noughty illusions contrast decreasing quantifiers such as few clouds and their increasing counterparts, a few clouds (v. op. cit. n. ). A project to eliminate what appears special to negation and to derive it instead from independent features of event semantics is incomplete without decreasing quantification. See Schein (: §) for argument that the contrast derives from a syntax and semantics distinguishing DPs with a zero article, few clouds, from those with ‘a(n)’, a few clouds. ¹⁸ The same is to be said for the pronouns qua definite description in ()–().

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

()

The air passengers felt it rage outside, in the seat of their pants.

()

[9e’: . . . rage[e’] . . . ] 9e( . . . feel[e, e’] . . . )

()

It raging outside tripped the “Fasten Seat Belts” light inside.

()

[9e’: . . . rage[e’] . . . ] 9e( . . . trip[e’, e] . . . )

The “obvious” theory founders however on negative event descriptions, battered between inferences of Perceptual Veridicality and Consistency (Barwise ; Barwise and Perry , ) and suspect negative events (Vlach ; Higginbotham , , ; Casati and Varzi  and references cited therein): (Veridicality) () The passengers felt it not be calm. () ‘ It was not calm. (Consistency) () The passengers felt it not be calm. () ‘ The passengers did not feel it be calm. ()

The passengers felt an event of it not being calm.

()

[9e’: ¬( . . . calm[e’] . . . )] 9e( . . . feel[e, e’] . . . ) () ⊬ (). () ⊬ ().

An event of it not being calm does not preclude others in which it is. Both may exist and both felt. Plainly, () entails neither () nor (), and the “obvious” theory is mistaken to confound () with (), translating them as (), if it is to validate Veridicality and Consistency. Note that this is a puzzle of negation and negative event description. Consistency is about the exportation of negation, and the “obvious” theory already validates Veridicality for positive event description, inference from feeling calm to there being calm. To stir up negative events, imagine a wet and windy storm s such that: ()

The storm tripped the “Fasten Seat Belts” light and iced the wings, and the passengers felt the storm.

Given (), ()–() are also true and yet ()–() false: ()

It not being calm tripped the “Fasten Seat Belt” light.

()

It not being dry iced the wings.

()

The passengers felt it not being calm.

()

F It not being dry tripped the “Fasten Seat Belts” light.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

    () ()



F It not being calm iced the wings. F The passengers felt it not being dry.

Since the storm s did all these things, the negative event descriptions in ()–() do not describe it. But, if it isn’t a not being dry or a not being calm, the storm certainly isn’t a being dry or a being calm either. Thus, the ‘not’ in ()–() violates Excluded Middle: ()

¬ being dry[s] & ¬ not being dry[s]

()

¬ being calm[s] & ¬ not being calm[s]

Moreover, if the “obvious” theory parses the false () and () as () and () to confound them with () and (), which describe the storm, the ‘not’ in () and () is not the ‘not’ in () and (): ()

An event that was not dry tripped the “Fasten Seat Belts” light.

()

[9e’: not( . . . dry[e’] . . . )] 9e( . . . trip[e’, e] . . . )

()

The passengers felt an event that was not dry.

() [9e’: not( . . . dry[e’] . . . )] 9e( . . . feel[e, e’] . . . ) The ‘not’ in ()–() derives a negative description ⌜not Φ[e]⌝ that does not denote all events that are not as it says they shouldn’t be, ⌜¬Φ[e]⌝. Those that are denoted, negative events, mentioned only by what they are not, are as forceful in every way the storm they are not is, tripping lights, icing wings, and tormenting passengers, ()–(). The ‘not’ in ()– () is then held to be other than ‘¬’, an antonymic operator non- (Higginbotham , , ; et al.) deriving descriptions of these negative events and violating Excluded Middle so that the storm s, not a negative event, is neither a being dry nor a being non-dry, (). A richer logical form for negative event descriptions in which negation is noughtly escapes ambiguity in ‘not’ and an occult meaning that invokes negative events and violates Excluded Middle (sections ..–..). It derives Perceptual Veridicality and Consistency (section .), too.

... The objects of perception Existential quantification in () over the objects of perception undermines Veridicality. Definite reference to it not being calm would restore that there is only one such. But, sideby-side could be the event of it not being calm and the event of it being calm. For Veridicality, these two must be mutually exclusive in any context for any true utterance of (). It must be explained how the objects of perception that populate a perceptual field conspire for it to contain it not being calm if and only if it not contain it being calm. Maybe it is their nature, as would be if the objects of perception were facts instead of events: there

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

is the fact of it not being calm if and only if there is not the fact of it being calm. It is also safe that the storm that is not calm is not the fact that it is not calm. Facts, like propositions or thoughts, come ready-made to host negation, free from the metaphysical dereliction of negative events. An important response (Barwise and Perry , ; Cooper ) to negative events, negative event descriptions, and the logic of negation in perceptual reports denies the event semantics in favor of reports about facts (formalized as situations σ that are truth-makers, σ ⊨ p, for both negated and unnegated p, v. Stojanovic , Zucchi ). As much as facts play nice with negation, a “Slingshot” argument (so-called in Barwise and Perry , , v. Davidson ; Neale , ; Neale and Dever ) dooms them as the objects of perception: ()

Galileo felt himself lose/losing his grip on the sphere.

()

Galileo felt the sphere land(ing) on the piazza  meters below.

Galileo remains turreted so that () is true and () false. A semantics that mistakes () and () for reports of the fact of Galileo losing his grip and the fact of the sphere landing on the piazza mistakenly proves () and () equivalent (v. Appendix: Gödel’s Slingshot). If the objects of perception are then not facts, it remains for event semantics to be reconciled to Veridicality and to face on its own terms the problems of negative event descriptions and negative events.

... Negative event descriptions Negative event descriptions join negative thing and negative stuff descriptions. Negative events happen, and negative things and negative stuff, too, linger: ()

A consular agent not giving us our visa was after another not saying hello and before another not saying goodbye. (After Horn (: ).)

()

A not yet a meringue plopped on the baking sheet.

()

Some not yet meringue spilled from the mixing bowl.

()

*[A(n) e: ¬(a consular agent not giving(e) us our visa)] *[A(n) x: ¬(yet a meringue(x))] *[Some x: ¬(yet meringue(x))]

Given the plenitude of events that are not a consular agent giving us our visa, not another not saying hello, and not another not saying goodbye, why isn’t () trivially true? Our storm s is none of what ()–() denote although it is as described, along with most anything else, if ()–() are parsed as in (), which are plainly then incorrect translations. Note that if ()–() were parsed as in (), with both noun and count/mass morphology in the scope of negation, the egg white would be both not-yet-meringue and not-yet-meringues:

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

    ()

Three not yet meringues plopped on the baking sheet.

()

*Three not yet meringue plopped on the baking sheet.

()

*Too much not yet meringues spilled from the mixing bowl.

()

Too much not yet meringue spilled from the mixing bowl.



That is, the negated NP ought to be a mass term, so that () ought to be as semantically anomalous as () (cf. *three egg white), and () ought to be as coherent as () (cf. too much egg white). Rather, count/mass morphology, like tense morphology, escapes its host to command a phrasal scope that includes negation. An affix, it still attaches to an unspoken NP, three relevant thing-s not yet meringue, too much relevant stuff-∅ not yet meringue: ()

[9X: three[X] & NP.[X] & [℩E: One-on-One[E,X]][No e: Ee] yet meringue[e]]

()

[9X: too much[X] & NP.∅[X] & [℩E: Coincide[E,X]][No e: Ee] yet meringue[e]]

A plenitude of negative events, things and stuff and occult negation are an illusion of the bad grammar in () that is blind to the descriptive content outside the scope of negation. Negation, noughtly, as restricted quantification, must find its restriction, which it does in ()–(). The count/mass morphology outside its scope betrays its presence, which also meets the requirements of a(n), three, and too much to modify count or mass NPs locally. The storm s isn’t yet a meringue but also not among the relevant things whose culinary preparation are at issue, one of which () says plopped on the baking sheet. Antonymy is an illusion of contextual and tacit restriction to relevant events, things, or stuff. There is implicit understanding of what is relevant, as when ‘relevant’ is spoken. Restriction to the relevant removes from the semantics of ()–(), (), and (), the invitation to negative events, things, and stuff¹⁹ and ambiguity in ‘not’.

... The scope of negation within descriptions What fixes the restriction to noughtly ‘[No e: Ee]’ varies with its scope and the antecedent content in the clause in which it occurs. Recognizing the scope effects resolves the ¹⁹ Metaphysical and semantic problems remain for (i)–(vi) but not for the negative events of (i) that merit solution separate from the other possibilia in (ii)–(vi): (i) (ii) (iii) (iv) (v) (vi)

the three naps that I did not take the two dream homes that I did not build the quart of meringue that never came together the three naps that I might yet take the two dream homes that I hope to and will possibly build the quart of meringue that more whipping is certain to whip up

See Condoravdi, Crouch, and van den Berg (), Condoravdi et al. (), and Varzi () for discussion and references therein.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

noted puzzles. The following lightly modifies van der Does’s () important, neglected demonstration of these scope effects. Imagine that the Traffic Department installs surveillance to assess the efficacy of red zones as a deterrent against illegal parking. There is no deterrence absent an intention toward the illegal behavior, and deterred intentions are often unobservable as maneuvers or hesitations in the traffic flow. The Traffic Department’s assessment is largely a statistical inference from the traffic flow recorded for a statistically significant period. Sentences ()–() report that for the period under surveillance, there was no parking in the red zone, the conditions tout court for () to be true: ()

Surveillance saw not a single motorist park in the red zone.

()

Surveillance saw no motorist park in the red zone.

()

Surveillance saw none of the motorists park in the red zone.

()

No motorist parked in the red zone.

In contrast, although it is sufficient for () that Sandy walk or speed by without noticing the red zone or even find herself across town, it is not so for () which demands of Sandy an omission that has saved her from the commission, an act of non-parking, again insinuating an occult meaning for ‘not’:²⁰ ()

Surveillance saw Sandy not park in the red zone.

()

Sandy did not park in the red zone.

Surveillance caught at least a deceleration in Sandy’s driving if not an aborted parking maneuver. Universal generalization in ()–() next demands omissions from commission from all the motorists, belying logical equivalence with ()–()²¹ and clamoring therefore for occult ‘not’ so long as the scope of ‘not’ in () and () is neglected: ()

Surveillance saw every (one of the) motorist(s) not park in the red zone.

()

Surveillance saw all ((of) the) motorists not park in the red zone.

²⁰ As in Higginbotham (), to see Sandy sleep is not to see her not leave, which is to see her stay. A not-leaving is a staying, and sleep isn’t a not-leaving or a staying. ²¹ A period of surveillance that saw a dead street and the red zone empty of any traffic verifies ()– () but falsifies ()–() absent drivers and their deterred behaviors, which shows that ()–() ⊭ ()–(). With a different context and examples, van der Does () shows in effect also that ()–() ⊭ ()–(), which is shown here adapting for this context. Suppose the motorists are some particular suspects under surveillance who have driven by the red zone outside their secret meeting place on multiple occasions. Sometimes they parked, stupidly in the red zone, and sometimes suspecting surveillance, they aborted parking. Now ()–() ⊭ ()–() in that ()–() are true with them all having aborted a parking and ()–() false on account of them all having also parked.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   



The scope effects belong to a class of such in which variation in the scope of not, no, never, few, and rarely, etc., varies the length of the antecedent description outside their scope and thus varies what motorist behavior is said to be surveilled and measured—what fraction of their intentions to park are deterred, as above, or what fraction of their parking was in red zones during rush hour, or what fraction of their parking in red zones was during rush hour: () Surveillance saw not a motorist park in red zones during rush hour. () Surveillance saw motorists not park in red zones during rush hour. () Surveillance saw motorists park not in red zones during rush hour. () Surveillance saw motorists park in red zones not during rush hour. () Surveillance saw no motorists park in red zones during rush hour. () Surveillance motorists park in no red zones during rush hour. () Surveillance saw motorists park in red zones during no rush hour. () () () ()

Surveillance saw rarely a motorist park in red zones during rush hour. Surveillance saw motorists rarely park in red zones during rush hour. Surveillance saw motorists park rarely in red zones during rush hour. Surveillance saw motorists park in red zones rarely during rush hour.

Sentences ()–() compare to ()–() in which spoken antecedent content fixes the reference of spoken descriptive event anaphora (v. () above): () Surveillance saw not in any of it a motorist park in red zones during rush hour. () Surveillance saw motorists act and in not any of it park in red zones during rush hour. () Surveillance saw motorists park and not any of it in red zones during rush hour. () Surveillance saw motorists park in red zones and not any of it during rush hour. () Surveillance saw rarely then a motorist park in red zones during rush hour. () Surveillance saw motorists act and rarely then park in red zones during rush hour. () Surveillance saw motorists park and rarely then in red zones during rush hour. () Surveillance saw motorists park in red zones and rarely then during rush hour. What fixes the restriction in ‘[No e: Ee]’ and ‘[Rarely e: Ee]’ in ()–() are unspoken counterparts to the descriptive anaphora in ()–(). Event semantics supplies unspoken thematic relations such as ‘Agent(e,x)’, counterpart to ‘act’ in () and (), so that surveillance captures action that is not parking, the motorists’ and Sandy’s omissions. Thus, ‘not’ as restricted adverb of quantification noughtly joins neo-Davidsonian decomposition (section .) and sub-atomic descriptive event anaphora (Schein ; ; :

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

ch.  §)²² to reduce to logical syntax the differences in meaning between ()–() and ()–() and among ()–(), without wanton metaphysics or occult vocabulary. Negative events and an ad hoc meaning for ‘not’ are an illusion that fixes its gaze on ‘not’ in ()–() blind to the rest of van der Does’ () paradigm from which () is excerpted.

.. V  

.................................................................................................................................. The “obvious” theory quantifies existentially over the objects of perception in () and forfeits Veridicality and Consistency, for which definite reference to it not being calm must crowd out from the perceptual field it being calm. ()

The passengers felt it not be calm.

Definite description by localized plural abstraction (section .), in agreement with nominal definite descriptions ()–() and the translation of descriptive anaphora ()–(), constricts reference in () to just those flight segments of a two-hour flight that witness turbulence, which means that the passengers can sleep through most of it when the clouds are rather scattered: ()

During their -hour flight, the passengers felt it not (be(ing)) calm underneath the clouds. But, it amounted to only  minutes under  clouds without a bathroom break.

Equally, definite reference to the objects of perception in ()–() constricts reference to just those who perceive it, agriculture department inspectors i, i and i from the morning and not i and i from the afternoon, just as definite reference refers to just the farmers, donkeys, -lb. bags of oats and  minutes of the two morning events (v. ()–()). () i

d

d >b – min (late am) d d f—d—b— min (early pm) i f—d—b— min (late pm)

f < i i

>b – min (early am)

i

f
b — min (early am) d f—d—b— min (early pm) d f< >b — min (early pm) d

i

i i

d f< >b — min (late am) d f—d—b — min (late pm) d f< >b — min (late pm) d

If so, then judging ()–() in context (), it must be that () ⊨ no more than two farmers each feed two donkeys one bag of oats, and yet () 6¼ the no more than two farmers each feeding two donkeys one bag of oats, the object of perception. Thus empirical considerations join “Slingshot” arguments (section .., Appendix) against the view that would make this identification in claiming that perceptual reports refer to facts (or to the situations of Situation Semantics). Reference to less than what it is true of obtains when the complement in a perceptual report is a definite event description derived by localized plural abstraction, which embeds the canonical logical form for negation () in the logical form for () (v. ()) to cull the turbulent flight segments: ()

[℩E: during -hr flight[E]], [℩E’: [No e: Ee & local[E,E’]] (calm[e] & [the Y: clouds[Y] & [8y: Yy]Xy & local[e,Y]] underneath[e,Y])] . . . ‘During their -hour flighti, where E’ theni, there was not a calm underneath the clouds . . . ’

The first sentence of () and ()–() all agree in what the passengers felt— minutes of turbulence: ()

During their -hour flight, the passengers felt when(ever) it was not calm underneath the clouds.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

()

During their -hour flight, the passengers felt the not being calm underneath the clouds.

()

During their -hour flight, the passenger felt whatever not being calm underneath the clouds there was.

But, suppose that the ten turbulent clouds are scattered among hundreds in the flight path under which it was calm, so that () is false while ()–() remain true, referring to the same  minutes of turbulence. In a general theory of definite description that runs from the demonstrative those clouds to the irrealis whatever clouds (if any), natural language descriptions supplement the common ℩-operator with content expressive of their variation in epistemic and existential commitments. If the complement to a perceptual report were to rely in its translation exclusively on meaning common to all definite descriptions, () would fail Veridicality in being true like ()–() when its complement is false. As a point of grammar, the event definite description that translates a perception verb complement includes realis or actuality morphology that says it is a true description, in appositive modification of its host:²³ ()

During their -hour flight, the passengers felt it not be calm underneath the clouds. ‘During their -hour flight, the passengers felt whatever, given it then not being calm underneath the clouds, it not being calm underneath the clouds there was.’

()

. . . feel Φ ) . . . [℩E’ : True(⌜Φ⌝). Φ[E’]] feel[E’]

That is, the perceptual report is Veridical because its complement clause says as much qua realis definite description. The Veridicality of perceptual reports, such as it is (cf. section ..), derives from the semantics of definite descriptions, for both unnegated and negative event descriptions, without need for a logic of perceptual reports to intervene on behalf of those of their complements that contain negation.

... Veridicality and consistency as cinema verité Veridicality, inference in the pattern () ‘ (), is easily dissolved, despite its appearance as intrinsic to the logic of the construction: ()

The passengers felt it not be calm.

()

‘ It was not calm.

²³ Inside definite descriptions, the syntax of modification is “layered,” showing positional and scope effects among a variety of modifiers, where leftmost is often reserved for appositives expressing a relation between the accumulated referents and their location in the context. See Schein (: §..; ).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   



A focus group is convened to select the final farewell scene in which the heroes face each other on a foggy night in Casablanca as a Lockheed Model  Electra Junior prepares for departure. On the split screen, ()

The focus group saw Rick clasp Ilsa’s hand, and it saw Rick not clasp Ilsa’s hand.

The scenes are simultaneous, as is the focus group’s experience of them, and yet no contradiction is inferred from the Veridicality of ()’s conjuncts. Actuality may replace artifice, with more background for the events in Casablanca. By prior arrangement, clasping none of Ilsa’s hands signals that Rick has no travel documents in his possession, one hand signals documents for one passenger, and both hands signals travel documents for two. Victor, the interested party, sees Rick clasp exactly one hand, of which it can be variously said in report of what happened that foggy night that: ()

a. Victor saw Rick clasp her hand and not clasp her hand. b. Victor saw Rick clasp her hand and Rick not clasp her hand.

()

a. Victor saw Rick clasp her hand and saw him not clasp her hand. b. Victor saw Rick clasp her hand, and he saw her not clasp her hand.

In apparent contradiction to the Veridicality of the second conjuncts in ()–(), it is not felicitous report of that night’s events that: ()

Rick did not clasp Ilsa’s hand.

That foggy night at the airport or in the cinema with the focus group, there are always spatiotemporal frames tacit or overt for the seeing, for the heroes’ embrace, and for the clasping or not of hands, etc.: () [℩E⁰: Φ⁰[E⁰]](Victor saw[E⁰] [[℩E: Φ[E]]Rick clasp her hand and [℩E: Φ[E]]Rick not clasp her hand] Then&there⁰, Victor saw Rick then&there clasp her hand and Rick then&therez not clasp her hand. ()

[℩E¹: Φ¹[E¹]](Victor saw[E¹] [[℩E: Φ[E]]Rick clasp her hand]), and [℩E²: Φ²[E²]](Victor saw[E²] [[℩E: Φ[E]]Rick not clasp her hand]) Then&there¹, Victor saw Rick then&there clasp her hand, and then&there², Victor saw Rick then&therez not clasp her hand.

()

[℩EC: Casablanca Airport  December [EC]] [℩x: Rick[x]][℩E: local[E, EC] & Agent[E,x]][No e: Ee](clasp Ilsa’s hand[e]) Casablanca Airport  December , Rick did not clasp Ilsa’s hand.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

The Veridicality of complement clauses delivers from () only that Rick clasped Ilsa’s hand then&there and not then&there, without contradiction provided that then&there 6¼ then&there. And, the Veridicality of () does not contradict () provided that then&there 6¼ Casablanca Airport  December . The “logic of perception” does not validate such Veridicality inferences as () ‘ (); and, as ()–() show, it is good that the syntax and semantics of perceptual reports and of negation are articulated enough to represent their failure. Their success is an illusion frequent enough to have been mistaken for a logic and merits further comment. First, there is likely a constructionspecific spatio-tense agreement that demands that then&there⁰ in () cover the elapsed spatiotemporal reference in the complement, so that then&there⁰ = then&there⌒ then&there, with the effect that () composes either a stationary wide-angle scene with two attentional foci or a medium-angle scan from then&there to then&there. It is not unexpected that such agreement constitute the tense dependency intrinsic to naked infinitival clauses.²⁴ Second, for any sequence, . . . then&thereij . . . then&therekl . . . , there are background, pragmatic assumptions that favor identity or continuity in spatiotemporal frames of reference, especially in the absence of explicit redirection, as a matter of discourse or narrative coherence, effects known from sequence of tense and discourse anaphora in general (v. Schein : ch.  §; Schein : ch. ; Appendix ). The grammar of tense dependency and the pragmatics of narrative continuity favor frequent default to the spatiotemporal coreference that makes Veridicality appear valid, as in () ‘ (). Perceptual Consistency is an illusion under like conditions, also dissolving in the above context in () ⊬ (), in contrast to () ‘ (): ²⁴ More is to be said about the conjunction embedded in () and in (iii): (i) Nelson Mandela heard the United Nations applaud. (ii) Nelson Mandela heard Afghanistan, Albania, . . . , Zambia and Zimbabwe applaud. (iii) Nelson Mandela heard Afghanistan applaud, Albania applaud, . . . , Zambia applaud and Zimbabwe applaud. (iv) Nelson Mandela heard Afghanistan applaud, (he) heard Albania applaud, . . . , (he) heard Zambia applaud, and (he) heard Zimbabwe applaud. Sentence (i) is unambiguously true, and (iv) false. It might be held that the semantics for (iii) ought to bake in a distributivity of perceptual reports over conjunction (Barwise ; Barwise and Perry , ; Cooper ), according to which (iii) entails (iv) and is therefore just as false. Let it be pointed out, as Higginbotham () does, that in the complement to a perception verb there are as many event descriptions as there are non-finite clauses, so that there is no daylight between (i)–(iv) and (v)–(viii): (v) Nelson Mandela heard the United Nations’ applause. (vi) Nelson Mandela heard Afghanistan’s, Albania’s, . . . , Zambia’s and Zimbabwe’s applause. (vii) Nelson Mandela heard Afghanistan’s applause, Albania’s applause, . . . , Zambia’s applause and Zimbabwe’s applause. (viii) Nelson Mandela heard Afghanistan’s applause, (he) heard Albania’s applause, . . . , (he) heard Zambia’s applause, and (he) heard Zimbabwe’s applause. The entailment—whether semantic or pragmatic—between (iii) and (iv) mirrors that between (vii) and (viii), the account of which will be the same in full support of the thesis that (iii) contains definite event descriptions just like (vii). Crucially, neither in (iii) nor () is the conjunction contained within a single event description, *the event(s) of Afghanistan applauding, . . . , and Zimbabwe applauding, *the event(s) of Rick clasping her hand and (Rick) not clasping her hand.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

    ()

The passengers felt it not be calm.

()

‘ The passengers did not feel it be calm.

()

The focus group/Victor saw Rick not clasp Ilsa’s hand.

()

⊬The focus group/Victor did not see Rick clasp Ilsa’s hand.



.. P 

.................................................................................................................................. Given that ‘not’ is noughtly, restricted quantification with the canonical logic form (), whether () is plain or embedded in a definite description of events, ⌜[℩E’ : ()]⌝, there are always the events E that Φ and those E’ of them that are as () describes, to which the definite description refers. ()

(Negation in Event Semantics) [9E: Φ[E]] [No e: Ee] Ψ[e]

Throughout ⌜[9E: Φ[E]]⌝ has stood for a natural language spatiotemporal frame adverbial, spoken or unspoken, which may itself impose constraints on what can substitute for Φ in an (in)definite description of (the events occupying) some spatiotemporal region. So much now aligns negative event descriptions and their non-negative counterparts so that their reference coincides: ()

Aloft for  hours, it not being calm tripped “ ” on for  minutes total and sickened twenty passengers.

()

Aloft for  hours, it being turbulent tripped “ ” on for  minutes total and sickened twenty passengers.

()

Aloft for  hours, it not being turbulent tripped “ ” off for  minutes total, leveled the aircraft, and calmed the passengers for about as long.

()

Aloft for  hours, it being calm tripped “ ” off for  minutes total, leveled the aircraft and calmed the passengers for about as long.

Assimilating () to () and () to () does not yet tame for the language of causation its problem of negative events, where a narrow gaze on () and () broods in despair at how nothing happening can be as effective as something happening for tripping a light switch. Across ()–(), event semantics sees invariance in a recurrent verb phrase, trip “Fasten Seatbelts” on/off, sicken/calm the passengers, or level the aircraft, and in ()–()

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

sink the ship—9e’(Cause(e, e’) & V(e’) & Theme (e’, s)). Joined to this invariance, it sees equivocation in the subjects’ thematic relations θ: ()

9e (θ(e, x) & 9e’(Cause(e, e’) & sink(e’) & Theme (e’, s)))

()

It not being empty of enemy traffic in wide-open seas sank the ship.

()

It being crowded with enemy submarines and narrow channels for escape sank the ship.

()

Loose lips sank the ship.

()

That the enemy knew its location sank the ship.

()

The defense minister sank the ship.

()

The submarine sank the ship.

()

The torpedo gunner sank the ship.

()

A weak hull sank the ship.

()

The torpedo sank the ship.

()

The torpedo explosion sank the ship.

If there is, for example, mental or intentional causation in this causal chain, ()–(), then ‘Content(e,x)’ is the relation between the causal mental or intentional state and its contents, v. (). Among these thematic relations must be a relation between the background conditions or spatiotemporal region that hosts such conditions and the causal events thereby enabled, v. ()–(), ()–(), and ().²⁵ It then remains an empirical problem to enumerate the thematic relations θ drawn upon in causal reports, which ought to reflect the underlying native theory of causation.²⁶ For now, it suffices to rescue talk about negative events from metaphysical mischief that the language of negation and the logical forms () are joined in parsing it.

²⁵ Varzi () looks at a spread similar to ()–() and concludes that only some are genuine causal reports (e.g. ()) and the rest are causal explanations (e.g. ()). It is however important that all tokens of sink the ship in ()–() be shown the same. It may just then be that the substitution of one θ for another in () turns causal report into causal explanation. ²⁶ For essential discussion of causal reports and their constituent thematic relations and concepts, Pietroski (, , ).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   



.. S

.................................................................................................................................. It is quickly discovered (section .) that negation in a language designed for a world of events as dense as space and time has the canonical logical form (), in which noughtly is a restricted, distributive event quantifier: ()

(Negation in Event Semantics) [9E: Φ[E]] [No e: Ee] Ψ[e] [℩E: Φ[E]] [No e: Ee] Ψ[e]

To understand negation is to have understood the E intended via spoken or unspoken (in) definite description or demonstration, which both circumscribes a spatiotemporal frame of reference and commits to some event segmentation within it. Varying with the scope of noughtly and similarly for rarely, mostly, etc. (section ..), unspoken descriptive anaphora to antecedently described events fix the E of its restriction, exploiting a logical language that conforms to the neo-Davidsonian separation thesis (section .) that all nominal arguments are arguments in their own, separate relation to events (section ..) separated from the adjectival or verbal predicates expressing monadic concepts of events (section ..). There is also a silent inventory (section .) of thematic relations, relations between events, ‘Cause(e,e’)’, and contextual restriction to the relevant events, things, or stuff (section ..). Consonant with the semantics of distributive quantification in natural language, definite descriptions in the scope of noughtly are localized definite descriptions restricted to the local event e (section .). For plural descriptions that quantify into clauses with distributive quantification in them, natural language is fitted with plural definite description by localized plural abstraction (section .), which constricts reference to what the local condition describes. Despite all the grammar, what is special to negation is just that ‘not’ is noughtly (but, v. n..). The rest of it finds its justification elsewhere. Negation in event semantics confronts the problem of negative events when negation is embedded in perceptual and causal reports as negative descriptions of events (section .). Crucial to their grammar is that the complement clause is a plural definite description of events that refers via localized abstraction to less than what it is true of (section .). The logical form for these constructions contain multiple references to spatiotemporal frames then&therekl. Coreference among them derives the inferences of Veridicality and Consistency (section ..), without arcane meaning for the perceptual report itself (section .., Appendix). Recognizing the scope effects that noughtly introduces when () is embedded eschews creatio ex ambiguous ‘not’ a plenitude of non-events, non-things, and non-stuff (section ..). The puzzles of meaning, logic, and metaphysics surveyed are an illusion of bad grammar. Their resolution is in the logical syntax of the language of thought, the language of event semantics in which ‘not’ is noughtly.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 



.................................................................................................................................. Gö’ S (N , ; N  D ) A bad choice for the semantics of perception verb complements ends in () entailing () when () is true: ()

Galileo lost his grip. Galileo’s sphere landed on the piazza. Galileo is not his sphere.

() Galileo felt himself lose/losing his grip. ()

Galileo felt his sphere land(ing) on the piazza.

Anyone who understands () and is competent with definite description and identity understands the equivalence between Galileo lost his grip and Galileo is the one who is Galileo and lost his grip and assents to the reasoning []–[] in (). Continuing []–[] with any sentence such as () embedding Galileo lose/losing his grip proves its counterpart (()) embedding Galileo’s sphere land(ing) on the piazza, unless the embedding context is found to be opaque to at least one of the substitutions in []–[]: () Gödel’s “Slingshot” (Neale : f.) [] Galileo lost his grip. [] Galileo is not his sphere. [] Galileo’s sphere landed on the piazza. [] Galileo is the one who is Galileo and lost his grip. [] Galileo is the one who is Galileo and not his sphere. [] Galileo’s sphere is the one that is his sphere and not Galileo. [] Galileo’s sphere is the one that is his sphere and landed on the piazza. [] The one who is Galileo and lost his grip is the one who is Galileo and not his sphere. [] The one that is Galileo’s sphere and landed on the piazza is the one that is his sphere and not Galileo. * * * * * * * * * * * * [] . . . Galileo lose/losing his grip . . . [] . . . Galileo be(ing) the one who is Galileo and lost his grip . . . [] . . . Galileo be(ing) the one who is Galileo and not his sphere . . . [] . . . Galileo be(ing) Galileo and not his sphere . . . [] . . . Galileo not be(ing) his sphere . . . [] . . . Galileo’s sphere be(ing) the one that is his sphere and not Galileo ... [] . . . Galileo’s sphere be(ing) the one that is his sphere and landed on the piazza . . . [] . . . Galileo’s sphere land(ing) on the piazza . . .

Premise Premise Premise ([], ℩-) ([], ℩-) ([], ℩- ) ([], ℩- ) ([,], ℩- ) ([,], ℩- )

Premise ([], ℩- ) ([,], ℩-) ([], ℩- ) ([], &-) ([], ℩- ) ([,], ℩- ) ([], ℩- )

Across the contexts that are uncontroversially event-denoting (Parsons ), []–[] mercifully founders at the start. The transient event of losing his grip is not the lifetime state of being Galileo, and while losing his grip is not while being Galileo (v. Sainsbury : f.):

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   



()

[] Galileo hurt just before he lost his grip. [] Galileo hurt just before he is the one who is Galileo and lost his grip. ... [] Galileo hurt just before his sphere landed on the piazza.

()

[] Galileo hurt during his losing his grip. [] Galileo hurt during his being the one who is Galileo and lost his grip. ... [] Galileo hurt during his sphere landing on the piazza.

()

[] Galileo hurt simultaneous to him losing his grip. [] Galileo hurt simultaneous to him being the one who is Galileo and lost his grip. ... [] Galileo hurt simultaneous to his sphere landing on the piazza.

()

[] Galileo hurt just before (the event or state of) (him/his) losing his grip. [] Galileo hurt just before (the event or state of) (him/his) being the one who is Galileo and lost his grip. ... [] Galileo hurt just before (the event or state of) his sphere(’s) landing on the piazza.

Just so, a syntax and semantics for () that designates the object of perception to be the event of Galileo losing his grip disarms the “Slingshot” at []–[]: ()

[] Galileo felt himself lose his grip. [] Galileo felt himself be the one who is Galileo and lost his grip. ... [] Galileo felt his sphere land on the piazza.

Event semantics is thus immune to the “Slingshot” but then must face negative event descriptions and negative events. As Neale (: f.) remarks in his magisterial work on the subject (, ; Neale and Dever ), the rejoinder that defenders of facts and situations (Barwise and Perry , ; et al.) make to other “Slingshot” arguments, denying the substitutivity of arbitrary logical equivalents, is no rejoinder to (), which substitutes only “Gödelian equivalents” like Galileo lost his grip and Galileo is the one who is Galileo and lost his grip. It is one thing that attention and inattention to Mercury may constitute distinct thoughts or information states for () and () despite logical equivalence and allow Galileo to hold them in different regard, making the context opaque to their substitution: () Galileo felt that the earth moves. ()

Galileo felt that the earth moves and an unseen planet perturbs Mercury or not.

() Galileo felt that the earth is the thing that is the earth and moves. It is another to forge a dissociation between () and (). If the object of the attitude is a situation s, a (partial) model for the language, s ⊨ p (Barwise and Perry , ), then () and () are reports of the same situation. If, say, instead, the objects of the attitudes are sentences, they are indeed

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 

distinct in () and (). But, only by reason of language deficit, Galileo’s or the speaker’s, is one of () or () asserted and the other denied. Either Galileo is incompetent in the language of definite description and identity and the speaker intends to convey as much, or the speaker is herself too incompetent to recognize that Galileo’s competence in his own language precludes contrary attitudes. “Slingshot” () shows that the objects of perception Galileo losing his grip and Galileo being the one who is Galileo and lost his grip cannot be the same, which, as event and state, they are not, as above. If they are instead facts, anything between situations and sentences, () shows that they cannot be the same fact, that is the fact of Galileo losing his grip is not the fact of Galileo being the one who is Galileo and lost his grip. Furthermore, with facts as the objects of perception, () shows that eagerly accepting () and rejecting (), as we all do, one ought find plausible that Galileo is in touch with the fact of his losing his grip but not with the fact of being he who is Galileo and losing his grip, which we do not. The same can be said about variations on facts, situations, states-of-affair, infons (Cooper ), etc. “Slingshot” () shows that the situation, state-of-affairs, or infon that Galileo lost his grip cannot be the situation, state-of-affairs, or infon that Galileo is the one who is Galileo and lost his grip. If the theory has not already ruled out the non-identity, it then begs the question how a rational agent such as Galileo could know or perceive the one and not the other.

A

.................................................................................................................................. Many thanks to Lucas Champollion, Terje Lohndal, and two anonymous reviewers for extensive comments.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  ......................................................................................................................

   Interaction with focus constituents ......................................................................................................................

 ˘ ˘ ̧

.. I

.................................................................................................................................. T relation between negation and alternatives constitutes a fruitful area of investigation for the semantics-pragmatics interface, with significant insights into how meaning is compositionally calculated and the role of context in this process. Negation interacts with a variety of constructions that have been argued to have a semantics based on alternatives, such as focus, questions, implicatures. This chapter will center on the interplay between negation and focus, which is affected by a host of syntactic, semantic, and pragmatic factors, and as such serves to illustrate the empirical issues and theoretical debates raised by the relation between negation and alternatives.¹ Alternatives, which are other expressions that could have been used in a given context, play a significant part in the grammaticality and felicity of utterances speakers produce and interpret. This led semantic and pragmatic theories to develop explicit models of alternatives and of their contribution to meaning, with applications to an increasing number of linguistic phenomena (see Fălăuș a for an overview of Alternative Semantics). The way alternatives are introduced or computed, as well as their interaction with operators such as negation, varies both across constructions and theories. For example, on some semantic theories (generally referred to as Hamblin semantics), denotations amount to alternative sets; these sets can be selected and operated on by sentential operators, for example quantifiers or negation (see Kratzer and Shimoyama , among many others). In contrast to this, on so-called ‘multi-dimensional’ theories of meaning (stemming from the analysis of focus developed in Rooth , ), alternatives, in particular focus alternatives, are computed separately from ordinary semantic values, with consequences

¹ The interaction between negation and questions will be dealt with in a different chapter (see Chapter ) in this volume and as a result, will be set aside in the present chapter.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 ˘ ˘̧

for the interaction of focus values and truth-conditional operators such as negation. Given the existence of these different theories, any discussion of the interaction between negation and alternatives or negation and focus needs to take into consideration these various ways of generating and computing alternatives. The notion of negation also needs to be qualified when investigating its interaction with focus and, more generally, with Information Structure (see also Chapter  in this volume). For most cases discussed in this chapter, we will be considering sentential negation, that is the kind of negation that syntactically combines with an entire clause and semantically applies to a proposition, changing its truth value. We will be less concerned with constituent negation occurring in examples such as Not everyone arrived late, where negation applies to a constituent smaller than a clause. A more relevant notion for the interaction between negation and focus is that of “contrastive negation” in sentences like John drank not coffee but tea. Karttunen and Peters () call this use of negation “contradiction negation,” Jacobs () and McCawley () use the term “contrastive negation,” and Jacobs () refers to it as “replacive negation.” Horn (, a) discusses instances of contrastive negation under the cover term “metalinguistic negation,” but we will set aside this latter notion as it will be treated in a different chapter (see Chapter  in this volume). Focus is not a uniform phenomenon either. There are multiple ways and theories to explain the role of focus in an utterance, but according to one prominent line of thinking focus evokes alternatives (see Krifka a; Beaver and Clark ; Büring  for comprehensive overviews). Informally speaking, “focus indicates the presence of alternatives that are relevant for the interpretation of linguistic expressions” (Krifka a: ). This broad definition applies to various focus marking strategies, including clefts such as It is London that she wants to visit or so-called in situ focus like She wants to visit [London]F, where the focused constituent is marked by pitch accent. The strategies used to mark focus and their interpretational effects may vary within and across languages. This definition does not say anything either about how the alternatives associated with focus are computed or whether their contribution to the overall interpretation is pragmatic or semantic in nature. As we will see shortly (section .), this is another point of contention in the literature on focus, which also determines how the interaction with negation is explained. Following the detailed overview in Krifka (a) on information structure, we can distinguish pragmatic uses of focus and semantic uses of focus. The two categories concern different aspects of meaning. Pragmatic uses relate to the communicative goals of the participants and infelicitous use of this type of focus leads to incoherent conversations. Semantic uses of focus concern factual information and affect truth conditions. Incorrect use of semantic focus may lead to unintended, possibly wrong, entailments. Standard examples of pragmatic uses include focus in answers to constituent questions, where the focused element has to correspond to the wh-element (). The question denotes a set of propositions (Hamblin ) and the answer serves to identify one of these propositions. In order for the answer to be congruent, the alternatives induced by the focused element must correspond to the propositions denoted by the question. In the case of so-called ‘presentational’ or ‘information’ focus (), the question can be implicit and can have broad, information-seeking forms like What happened?: () A: Who arrived late?

B: [Charlie]F arrived late.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   ()

And then something strange happened.



[A meterorite fell down]F.

Another prototypical case of pragmatic use of focus is to correct or confirm information, as in (), where the focus alternatives induced by the answers in B–B0 must include a proposition that has been proposed in the immediately preceding context (Krifka a: ): () A: Mary stole the cookie. B: (No,) [PEter]F stole the cookie!

B0 : Yes, [MAry]F stole the cookie.

Semantic uses of focus concern operators whose interpretation depends on focus and are said to be associated to focus. Prototypical examples are focus-sensitive particles like only, also, and even (discussed in sections . and .), which convey something about how the assertion relates to its alternatives. For example, in the case of exclusive particles like only in a sentence like Only Charlie arrived late, the prejacent (Charlie arrived late) is the only one among the alternatives that leads to a true assertion (i.e. no other contextually relevant individual arrived late). Additive particles like also or too convey that the assertion holds for other alternatives, for example a sentence like Martin also arrived late asserts that Martin arrived late and presupposes that some other alternative must be true (i.e. some other contextually relevant individual arrived late). Semantic effects of focus have also been noted with certain adnominal quantifiers in sentences like Few [incompetent]F cooks applied, where the use of focus leads to the inference that few of the cooks who applied were incompetent (see Partee ; Herburger  on the interaction between focus and quantification). There are other notions of focus that have been used in the literature and that qualify the kind of focus alternatives evoke or the way these alternatives are used. One dimension of variation that is relevant for the interaction with negation concerns the size of the constituent to which focus applies: full clauses (as in () above) or subconstituents such as VPs or DPs (). The notions of broad and narrow focus have sometimes been used for this distinction, especially in the syntactic literature (see Lambrecht ). The size of the set of alternatives can also vary, from very large sets to smaller sets, with the extreme case where the set of alternatives only includes two propositions, as in the case of corrective focus () or answers to constituent questions such as (): () A: What do you want to drink, tea or coffee?

B: I want [tea]F.

As already mentioned, some uses of focus are contrastive, that is depend on the existence of a proposition in the context with which the utterance can be contrasted (see Jacobs ). This can be easily observed in the case of corrective uses of focus, but it can also occur with additive uses, as in A: Charlie arrived late. B: [Sam]F arrived late, too. These brief remarks show that the interaction between focus alternatives and negation is a multifaceted area of investigation. Given the multitude and diversity of the interfering factors, it is clear that an exhaustive treatment of this complex matter goes beyond the scope of the present chapter. Instead, the following sections will focus on some prominent phenomena at the semantics-pragmatics interface where the existing contributions and the

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 ˘ ˘̧

remaining issues can be more clearly articulated. The chapter is organized as follows: Section . discusses the focus-sensitivity of negation and syntactic factors affecting the interaction between negation and focus. Section . investigates the role of negation in the distribution and interpretation of focus-sensitive particles. Section . is devoted to focused and non-focused polarity-sensitive items, elements whose distribution depends on the presence of negation and whose interpretation has often been argued to be tied to focus alternatives; section . offers some concluding remarks.

. N   

..................................................................................................................................

... Negation as a focus-sensitive element and contrastive negation It is well established that focus affects the interpretation of a sentence. One way to capture the meaning differences it induces is to assume that the function of focus is to evoke alternatives (an intuition that goes back to Jackendoff ). Rooth (, ) developed an influential compositional alternative-based semantics for focus, which assumes a ‘multidimensional’ semantics, with two levels of interpretation—the ordinary level and the level of alternatives: alongside standard meanings, speakers recursively build up alternative sets that are accessed by alternative-sensitive operators. On this theory, natural language expressions have two semantic values: an ordinary semantic value and a secondary, focus semantic value and, moreover, focus alternatives are introduced and computed separately from the regular semantic values. The focus semantic values are relevant for a variety of expressions in language, expressions which associate with focus, for example focus-sensitive particles such as only, also, or even. In Rooth’s system, they operate over focus alternatives and incorporate them into the ordinary meaning of the sentences where they occur. Different positions of focus determine different alternative sets, which once combined with the lexical semantics of a focussensitive expression, result in different truth conditions. Consider the following examples: ()

a. John only introduced Bill to [Sue]F. b. John only introduced [Bill]F to Sue.

The ordinary value of the two sentences is identical. Their focus value differs though: the focus value contains propositions of the form John introduced Bill to x for (a), and propositions of the form John introduced x to Sue for (b), where x is a variable of type e. Once only is computed, the resulting meaning of (a) is that ‘John introduced Bill to Sue’ and any other proposition in the set of alternatives is false, that is John introduced Bill to Sue and to no one else. In (b), only takes a different alternative set and the derived interpretation is ‘John introduced Bill and no one else to Sue’. This provides a simple illustration of the truth-conditional effect of focus-induced alternatives.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



Jackendoff () argues that sentential negation can also (optionally) act as a focus-sensitive operator, similarly to only. The discussion revolves around the association between negation and focus in sentences such as (), taken from Büring (: ): ()

a. Trane didn’t mention your [name]F in court. b. Trane didn’t mention [your]F name in court.

The intuition is that although their regular semantic value is identical, the examples in () end up differing in meaning due to the presence of focused constituents in the scope of negation. Accordingly, (a) negates that the addressee’s name was mentioned in court and implies that something else was mentioned, and (b) negates that the addressee’s name was mentioned in court and suggests that someone else’s name was. The question is whether these different inferences are truth-conditional effects or should be considered a pragmatic phenomenon. This issue has been controversial (e.g. Kratzer ; Jacobs ; Partee ; Rooth ; McCawley a; Horn a; Krifka a; Wagner ; Beaver and Clark ; Büring b, ), but the current consensus seems to favor a pragmatic solution, whereby there is no conventionalized association with focus for negation. One way to see this is the acceptability of continuations such as In fact, why should he involve you at all? or In fact, he didn’t mention you at all for (a). Since no contradiction arises, the sentence in (a) cannot entail that there is something other than the addressee’s name that Trane mentioned, suggesting that we are dealing with a weaker, cancellable, non-truth-conditional inference. In other words, focus is used to evoke alternatives, but unlike in the case of other alternative-sensitive operators, there is no commitment to any of the alternatives being true. While in the examples with only above we get an existential inference that John introduced Bill to someone (a), no such inference is obligatory in (). Rather, the sentence simply suggests that the choice among the focus alternatives of the form Trane mentioned your x in court is under discussion or salient. This effect can be derived from the semantics and pragmatics of focus alone in various ways, including by making reference to a representation where negation takes scope over the focus interpretation operator and does not interact with it (Hajičová ; Rooth ; Herburger ), or to the congruence between the set of alternatives and the question under discussion (Beaver and Clark ).² Details aside, although variation in focus placement in negated sentences produces different inferences, negation does not seem to manifest any kind of conventionalized focussensitivity.

... Negation and focused constituents: Syntactic factors The interpretive differences mentioned above have also been connected with (and attributed to) the syntactic representation of focus. A lot of research into the interaction between negation and focused constituents has centered on scope issues and evidence has accumulated that syntactic factors determine possible scope configurations. It has been argued that ² On the parallelism between focus and negation as cues for retrieving the question under discussion as well as the role of the question under discussion in negative sentence processing, see Chapter  in this volume and Tian and Breheny (). Thanks to a reviewer for pointing out the relevance of this study.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 ˘ ˘̧

dedicated focus projections can appear above and below negation and the various focused constituents can target these projections either at surface structure or in the Logical Form. The details of existing proposals along these lines vary with the language(s) under investigation. As an illustration, let us consider the following examples in Spanish, from Etxepare and Uribe-Etxebarria (): ()

No ha venido Pedro. a. No ha venido Pedro, sino que se ha ido María. not has come Pedro but that  has left María ‘It is not the case that Pedro has come, but rather that María has left.’ b. No ha venido Pedro, sino María. not has come Pedro but María ‘The one who came isn’t Pedro, but María.’ c. No ha venido Pedro, y no María. not has come Pedro and not María ‘The one who hasn’t come is Pedro, and not María.’

As the paraphrases show, the negative sentence in (), with a postverbal subject, allows three different interpretations. On one possible reading, illustrated in (a), negation takes scope over the whole clause, as confirmed by the continuation with a contrastive tag where the alternative involves a full clause (preceded by the complementizer que). In the second reading, in (b), negation takes scope over the postverbal subject, which is focalized. The continuation sino María (‘but María’), which is only possible for constituents under the scope of negation (not X but Y), confirms the narrow scope of the negative marker. Finally, in the construal in (c), negation applies to the VP and the focused postverbal subject outscopes negation (note however that there seems to be some variation among Spanish speakers concerning this judgment). This is shown by the acceptability of the continuation y no María (‘and not Maria’), which is only possible with positive constituents (X and not Y). According to Etxepare and Uribe-Etxebarria, these facts can be derived if we assume that there is a different structure involved in each construal. Based on facts pertaining to focus in Spanish and Basque, they argue that the semantic scope of focus derives from its syntactic scope and the different interpretations of focus in negative sentences follow from differences in the syntactic structures associated with their logical form. Specifically, they provide arguments showing that the narrow scope reading of focus (in b) correlates with a low focus projection, situated below negation and above the verbal projection, and the wide scope reading of focus, sometimes called free focus (c) results from movement of the focused constituent to a higher focus projection.³ Finally, the reading where the entire sentence is in focus (in a) involves no movement to a focus projection.

³ Note that a wide scope analysis might still be tenable if we assume a conjunction reduction analysis. In particular, the presence of the two negative markers in (7c) might favor such an analysis. In order to rule out a conjunction reduction analysis, one needs to consider sentences similar to (7) with scope bearing expressions in addition to negation and coordination. The prediction is that negation should take scope below such additional expressions, but we will not pursue this matter here. I am grateful to an anonymous reviewer for discussion of this point.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



The various interpretations of () and the syntactic account proposed by Etxepare and Uribe-Etxebarria illustrate the interplay between syntactic and semantic scope in sentences containing negation. Similar examples in Italian are discussed at length in Samek-Lodovici (), among others. Further possible connections between syntactic structure and focus in negative statements can be observed in the use of focus markers functioning as negative reinforcers, such as no in Northern Italian dialects, in () below: () a. Non ci vado NO! not there go  b. NO che non ci vado!  that not there go ‘I won’t go there’

Regional Italian Regional Italian

In these sentences, discussed in Poletto (), the marker no serves as a sentence final or sentence initial focus marker emphasizing the sentential negation non. This type of ‘doubling’ focus negative marker can over time turn into standard negation, as documented by Zanuttini () for certain Northern Italian dialects (e.g. Pavese and Milanese), for reasons pertaining both to information structure and syntactic structure. Poletto further shows how the various uses of this emphatic negation marker correlate with different positions of focus in the left periphery. Space limitations preclude an in-depth discussion of the Italian data and the proposed account, but these kinds of facts provide morpho-syntactic and semantic evidence for the close connection between negation and focus markers and the importance played by syntactic factors in their use. For further cross-linguistic evidence and discussion of the relation between syntactic vs. semantic scope of negation with respect to focus structure, see Schwenter () on Brazilian Portuguese, Kaiser () for Finnish, Kiss () and Kenesei () for Hungarian, Neeleman and Vermeulen (), de Clercq (), among many others (see also Chapters  and  in this volume).

.. N   

.................................................................................................................................. Focus-sensitive particles like only and even, and their very diverse cross-linguistic counterparts, constitute one of the richest areas of investigation in formal semantics and pragmatics (e.g. Beaver and Clark ; Gast and van der Auwera ; König ). As mentioned in section ., only or even operate on a set of alternatives and express something about the way in which these alternatives are to be factored into meaning. While exclusives like only negate all alternatives non-entailed by the assertion, scalar expressions like even convey that the denotation of the focused element is ordered with respect to the other alternatives. The scalar component of the latter type of focus-sensitive particles makes the connection with negation particularly interesting, given the well-known scale-reversal effect of negation (a property discussed in the literature since Horn  and Fauconnier a).⁴ To see the ⁴ Lexical or contextual scales are typically ordered in terms of entailment (see Israel a; Fox and Katzir ; Fălăuș a; Coppock and Beaver  and references therein for extensive discussion). In

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 ˘ ˘̧

different behavior of scalar focus particles in positive and negative contexts, let us take a closer look at the meaning of even: ()

Charlie even read [Syntactic Structures]F. ) (Charlie read Syntactic Structures) likely (Charlie read x: x is a relevant book)

In (), the presence of even triggers the opposite inference, namely that Charlie reading Syntactic Structures is more likely than reading any other (contextually relevant) book. Given what we saw for (), this is unexpected: since negation is a hole for presuppositions (in the sense of Karttunen ), the presupposition that Syntactic Structures is less likely than its alternatives should be present in () as well, but instead we get the opposite inference. There are two main approaches to the behavior of even in negative contexts. On the lexical ambiguity theory, proposed by Rooth () (see also Rullmann , ; Giannakidou ; Erlewine , among others), there are two lexical items even with distinct interpretations: ()

a. [[even]] (C,p,w) is defined if and only if ∀q ∈ C [q ≠ p → p Reason] It is not because he lied that he is in jail.

()

A: He is in (some fucking) jail because he lied. B: He isn’t in (some fucking) jail because he lied. (That doesn’t add up.)

Horn (: ), observes that no natural language seems to lexicalize the distinction between external and internal negation. However, Kroeger () argues on the basis of extensive empirical evidence (i.e. syntactic distribution, and semantic scope) that in Malay/Indonesian internal (predicate) negation is expressed by the “standard” negation marker tidak whereas external (sentential) negation is expressed by the “special” negation marker bukan.¹¹ Relevantly, it is bukan that normally expresses metalinguistic negation, as illustrated by the sentences in ()–(), where “replacing bukan with tidak forces a shift from metalinguistic to descriptive negation, resulting in a logical contradiction: a person cannot buy six of something if he does not buy one, and a person cannot grow flowers for a living without planting flowers” (Kroeger : –). ()

aku beli  buah sekaligus! a. Aku bukan beli satu tau,   buy one know  buy six  at.once ‘I didn’t buy one, you know, I bought six at one time!’ b. #Aku tidak beli satu, aku beli  buah.   buy one  buy six 

()

a. Dia bukan menanam kembang tetapi bertanam kembang.   .plant flower but .plant flower ‘He doesn’t (just) plant flowers but grows flowers (for a living).’ b. #Dia tidak menanam kembang tetapi bertanam kembang.   .plant flower but .plant flower

¹¹ Cf. Oseki (), who also argues that Japanese nai and no-de-wa-nai lexicalize the distinction between internal (nai) and external (no-de-wa-nai) negation; see also Lee (, ) for Korean, and Mughazy () for Egyptian Arabic, where the metalinguistic negator mes can co-occur with the discontinuous descriptive negator ma-[verb]-(e)ʃ, as in (i) below, taken from Mughazy (: ). (i)

abu:-h father-his

meʃ 

ma-falsaʕ-ʃ -kicked.the.bucket-

lessa–huwwa yet–he

ma-twaffa:-ʃ -passed.away-

‘It is not that his father didn’t kick the bucket yet. He didn’t pass away yet.’

lessa yet

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 



() Perdana Menteri bukan mempunyai satu atau dua mata tetapi Prime Minister  has one or two eyes but mempunyai beribu-ribu mata . . . has thousands eyes ‘The Prime Minister does not have one or two eyes but has thousands of eyes . . . ’ (Kroeger : –) In addition to scope interpretation and lexical choice of negation marker, word order (i.e. syntactic distribution) also reveals the external nature of metalinguistic negation, especially when it is expressed by the type of peripheral idioms presented in section ... Studying the acquisition of negation by English children, Drozd () analyzed early child productions with sentence-initial no as tokens of metalinguistic negation, parallel to adult sentences with my eye and similar MN markers, and proposed for both the syntactic structure in (). () a. [CP No [IP Mommy doing ] ] b. [CP My eye [IP Al and Hilary are married ] ]

(cf. Drozd : )

The high structural position of No and My eye accounts for scope interpretation and word order (further IP-topicalization may derive the order Al and Hilary are married my eye). Besides, CP articulates the IP-layer propositional content with the discourse context. Though quite diverse, later works (by other authors) dealing with syntactic aspects of metalinguistic negation converge in taking the CP domain, also known as ‘left periphery’ (cf. Rizzi , among others), to be central to the syntactic encoding of metalinguistic negation (cf. Giannakidou and Stravou ; Giannakidou and Yoon ; Martins , ; Oseki ; Larrivé ). Topic, Focus, Emphatic/Expressive/Evaluative, as a kind of illocutionary force, have been referred to in the literature as activated layers of the left periphery in MN sentences. Figure . represents Rizzi’s (, ) cartography of the left periphery: “a system of functional heads and their projections . . . delimited upward by Force, the head expressing the clausal typing . . . ” (Rizzi : ). Rizzi’s left periphery has been enriched by further splitting some of the categories of the CP space, but also by adding pragmatically motivated structure above CP (cf. Speas and Tenny ; Haegeman and Hill ; Haegeman ; Hill ; Corr ). This is exemplified in Figure .. Corr’s () expanded left periphery incorporates an UtteranceP (UP) space above the CP space, where the Speech Act (SA) projections of Speas and Tenny () belong. Elements that activate (layers of) this extended left periphery “mediate between the propositional content encoded lower down in the functional structure—i.e. the at issue, semantic content that the speaker wants to communicate to their addressee(s)—and the extra-sentential context, i.e. the wider, interactive discourse environment into which, via the act of the utterance itself, the proposition is introduced” (Corr : ). Moreover, ‘clausal typing’ may be decoupled from ‘illocutionary force’, splitting Force into three different categories, as represented in Figure .. This is not the place to work out a detailed analysis of the syntax of metalinguistic negation, though it is expected to display a lot of microparametric variation in view of

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



   LEFT PERIPHERY (a.k.a C-system, CP space) Force Top* Int Top* Foc Mod* Top* Fin Discourse-oriented layer

IP … Propositional layer

 .. Rizzi’s (, ) left periphery (Top – Topic, Int – Interrogative, Foc -Focus, Mod – (Adverbial) Modifier, Fin – Finiteness)

LEFT PERIPHERY (UP space above CP space)

[SAHighP [SALowP [EvalP [EvidP [DeclP [TopP [Pol-IntP [ExclP [Wh-IntP [FocusP [FinP [IP … Split ForceP

Syntactic encoding of ‘conversational dynamics’

 .. Corr’s () left periphery (SA – Speech Act, Eval – Evaluative, Evid – Evidential, Decl – Declarative, Pol-Int – Polar-Interrogative, Excl – Exclamative, Wh-Int – wh-Interrogative)

the diversity of MN markers available across languages and within the same language (see section .). But maybe Figures . and . can give a sense of the possibilities that current generative syntactic models (especially under the cartographic approach) open to a better understanding of the syntax of metalinguistic negation. In the remainder of this section, I will briefly comment on four aspects of the syntax of MN that signal its close association with the clausal left periphery. (a) Among the unambiguous MN markers identified in section . are wh- phrases, which are elements of the CP space, a matter of general consensus. The fact that wh- MN markers may surface in final position in languages with overt wh-movement (see () above) thus shows that the IP-part of the clause can be moved to the left periphery as well, presumably to a Topic position, as it corresponds to the objected given information (cf. Martins ; Larrivé ). So, it is expected that other MN markers may alternate between sentence-initial and sentence-final position, as is usually the case with idioms and swear words (see ..). The availability of EP sentences like (Bd) with the sentence (i.e. the IP constituent) sandwiched between two peripheral wh- phrases indicates that there is more than one position in the left periphery for MN markers to be merged (either by external or internal merge), which is a possible source of microparametric variation. (b) Other MN markers display the complementizer que, syntactizing the articulation between an idiom (cf. ..) or other kind of item (cf. ..) and the rest of the sentence. Romance que is the head of the original unsplit CP. But as Corr’s () fine-grained analysis of the left periphery of Ibero-Romance languages has convincingly shown, que can in fact be associated with different functional positions in the UP and CP space, each mapping into a specific grammatical/pragmatic value. The relevant point here is that que is

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 



always the spell-out of some functional category of the left periphery. Similar evidence is provided by Egyptian Arabic, where a negative matrix clause can take an overt complementizer only if it expresses MN, not DN (Mughazy ). (c) It has been observed that metalinguistic negation is a Root/Main Clause phenomenon (cf. Horn ; Martins ), thus typically occurring in main clauses, but possible in some types of embedded clauses and excluded from others.¹² While some authors account for the distributional restrictions of main clause phenomena (MCP) in pragmatic terms (e.g. based on the intuition that MCP are licensed only in clauses with a certain degree of illocutionary independence—cf. Hooper and Thompson ; Bentzen ), others analyze MCP in syntactic terms, though in relation to pragmatic notions such as specification for point of view (Speas and Tenny ) or anchoring to the speaker (Haegeman ). Under syntactic accounts, what excludes MCP from certain kinds of subordinate clauses is their impoverished left periphery, which lacks certain functional layers required by Root constructions. Hence, the Root nature of MN suggests that it requires full activation of the left periphery, including its higher utterance-oriented layers. (d) The way MN interacts with polarity items is one of its defining properties, in contrast to DN. The non-licensing of NPIs may have two sources: (i) in MN sentences the Pol(arity)-head within IP does not have neg-features (Martins ); (ii) the MN marker sitting in the left periphery is not local enough to license NPIs belonging to the IP domain (Oseki ). These lines of reasoning are more easily implemented for sentences displaying unambiguous MN markers (which are either merged IP-internally and then moved to the left periphery or directly merged there) than for not-sentences. Further investigation is needed to determine whether it is possible and desirable to unify the syntax of the two types of MN sentences. If MN not-sentences can be shown to be structurally different from DN not-sentences (despite superficial appearances), Horn’s “pragmatic ambiguity” may turn up as a trivial case of structural ambiguity, with the MN/DN split resulting from where in clause structure and how the negative operator is interpreted in relation to which functional categories and their silent featural content. The interim conclusion is that, differently from DN, MN necessarily activates functional categories in the extended left periphery, and MN markers do not/may not bear a syntactically active [neg] feature.

.. C 

.................................................................................................................................. Recent experimental investigation on the issue of the “pragmatic ambiguity” of negation (ERP and eye tracking) did not scratch the polemic potential of Horn’s proposal, as experimental results go in opposite directions keeping all the appeal of the debate: Lee’s (, ) ERP experiments support the “pragmatic ambiguity” hypothesis; Noh, Choo, and Koh’s () and Blochowiak and Grisot’s () eye-tracking (and elicitation) experiments argue against it. Further experimental work will presumably come up assessing this and other theoretical hypotheses on the topic of metalinguistic negation, whose extensive ¹² See Kroeger (: –) for a brief review of the literature on the syntactic distribution of Root constructions.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



  

literature provides plenty of testing material. Language acquisition is another fertile ground for testing specific theoretical hypotheses and assessing cognitive aspects of metalinguistic negation (cf. Drozd ; Nedwick ). On the other hand, the broadening of focus from MN not-sentences to MN sentences in general, including those displaying unambiguous MN markers, has created an area with considerable potential for further research and explorations.

A

.................................................................................................................................. I would like to thank two anonymous reviewers for their valuable comments and suggestions, and the volume editors, M.Teresa Espinal and Viviane Déprez, for their work and guidance throughout the process of writing and publishing this chapter. The research presented here was partially supported by FCT—Fundação para a Ciência e a Tecnologia (Projects UID/LIN//; UIDB//).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  ......................................................................................................................

   ......................................................................................................................

    

.. N    ,   ’

..................................................................................................................................

... Projection and cancellation U the surface of the language that we use lies a murky substratum of meaning. This dark underworld contains extra implications which, while sometimes crucial to establishing what the speaker is trying to say, are not the main point the speaker intends, but rather assumptions that the speaker is making, presuppositions¹ as they are standardly termed. Keeping presuppositions backgrounded is an efficient way to structure information, and allows interlocutors to focus on that main point, the information that is atissue. Our communicative goals typically center on presenting and commenting on the status of this at-issue material, so when we negate a sentence we typically aim to negate at-issue material. But that suggests a question: how can hearers identify exactly what material a negation is intended to target, and, more specifically, when does negation target presuppositions? Over a half century of contemporary scholarship has shown that what we might term the deniability of different types of inferences varies as the nature of the inference varies. Whereas entailments have a tight relationship to the ordinary semantic content of an utterance, and are difficult to deny without using an explicit negation, Gricean conversational implicatures can typically be easily defeated; witness the difference between ‘It’s warm, in fact hot’ (denying the implicature of ‘It’s warm’ that it’s not hot) and ‘#It’s warm, in fact cold’ (an attempt to deny the entailment of ‘It’s warm’ that it’s not cold). Somewhere in between entailment and conversational implicature lie presuppositions, notorious for

¹ Readers unfamiliar with presupposition might find the following general overviews helpful: Potts (), Atlas (), and Beaver ().

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



    

their complicated relationship with negation. For example, let’s look at the presupposition trigger stop in (a). ()

a. b. c. d.

Danny stopped smoking Danny does not smoke now Danny smoked in the past #Danny stopped smoking. In fact, he never smoked in the first place!

Here, the use of the word stop triggers not only the inference in (b) that Danny does not currently smoke, but also the presupposition in (c) that Danny did smoke at some point in the past. This inference cannot easily be canceled, in the sense that simply asserting (a) but then denying the inference in (c) produces an infelicitous discourse, as in (d). So, what happens to the inference in (b) when the sentence in (a) is negated, as in (a)? ()

a. b. c. d. e. f.

Danny didn’t stop smoking Danny smokes now Danny smoked in the past Danny didn’t smoke in the past Danny does not smoke now Danny didn’t stop smoking—he never smoked in the first place!

The most obvious way to interpret (a) is that the negation is targeting the at-issue semantic content concerning Danny’s current smoking habits, such that (b) now follows. However, the presupposition is unaffected by the negation, insofar as a hearer of (a) might still judge the inference (c), that Danny smoked previously, as valid. In such a case, the presupposition is said to project through the negation embedding. Projection is a signature property of presuppositions and has often been used as a diagnostic to distinguish them from other types of inferences like implicatures and entailment. Thus, it is the fact that the inference in (c/c) survives even when the host sentence in (a) is negated that constitutes the primary empirical reason for identifying that inference as a presupposition. Now things get a little more complicated: we can also get a reading of (a) that targets the presupposition instead of targeting the at-issue content alone, leaving us with the inferences (d–e). Here, the presupposition is canceled. On its own, this interpretation might be difficult to get. We seem to need the negation to be combined with an explicit denial like in (f), as if the premise to a joke was set up and we’re awaiting the punchline.² Along with this additional clarifying clause, we might also expect a shift in prosody, focusing the listener on the trigger stop rather than the negation in didn’t. Summarizing the data from simple positive and negated sentences: (i) entailments cannot be canceled without negation, (ii) implicatures can be canceled without negation, while (iii) presuppositions, like entailments, are not canceled without negation, but they also aren’t ordinarily targeted by negation, and are canceled when there is both negation and additional material denying the presuppositional inference.

² In contrast, (1d) is not merely joke-like, but strikes us as quite surreal.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



Although the listener might need extra help to interpret (a) as a presupposition denial, it is clear that the negation in (a) can also target the presupposition. So, we’re left with a puzzle: how do we account for the fact that presuppositions seem to be able to project and be cancelled by the same morpho-syntactic form? Is the negation in (a) ambiguous between a presupposition preserving (preservative) and presupposition cancelling (annihilative) function? If so, what might an annihilative negation operator look like? Describing this annihilative mechanism can have profound implications for the way that we define presuppositions in the first place. The conditions under which a presupposition is triggered and the conditions under which it fails to be triggered or is cancelled have traditionally provided the necessary data to define presupposition independently from other types of backgrounded information, like conventional implicature and entailment. So, as the nature of presupposition projection and cancellation has remained controversial, there has also been speculation over whether the constructions that have been identified as presupposition triggers actually form a distinct class.³ A brief terminological digression before moving forward: it is helpful to separate the terms negation, denial and cancellation. Roughly, we will use negation to indicate a morpho-syntactic device, the sort of syntactically realized operator or structure that is used to invert some aspect of meaning, typically truth conditions. We will reserve the term denial for the speech act of objecting to something in the discourse, whether that involves an explicit negation or not. Cancellation will be used broadly as an umbrella term for any instance where some change leads to an inference being removed. In principle, such removal is independent of whether morpho-syntactic negation is present, and independent of whether there was an explicit denial. For example, in ‘David ate some of the cake, and if he ate all of it he’ll be feeling sick right now,’ the implicature from ‘David ate some of the cake,’ that he didn’t eat all of it, is, in our terminology, cancelled, despite the absence of negation and the absence of any indication that a speech act of denial is being performed.⁴ Note further that what we refer to as annihilative negation has sometimes in the literature been termed p(resupposition)-canceling negation, metalinguistic negation, presupposition denying negation, or external negation, this latter partly on the basis that the ‘it is not the case that . . . ’ construction in English, where the negation is syntactically external to the targeted clause, has been thought to be associated with cancellation of presuppositions. Likewise, what we term preservative negation has sometimes in the literature been termed p-preserving negation, ordinary negation, descriptive negation, or internal negation. In our terminology, it becomes an empirical question whether syntactically external negation is indeed annihilative, and a separate empirical question whether clause internal negation is preservative. However, it is immediately clear that if there is any such alignment, it is at most a matter of preferred interpretation: presuppositions sometimes project through clause external negation, and sometimes fail to project through clause internal negation.⁵ ³ See e.g. Boër and Lycan ; Karttunen and Peters ; Lycan ; Burton-Roberts . ⁴ Horn () refers to cases lacking explicit acts of denial as involving suspension rather than cancellation. Here, instead of being denied, the truth value of the presupposition is left unresolved. ⁵ Presupposition annihilation negation is represented in the literature using both ‘It is not true that . . . ’ and It is not the case that . . . ’ (see e.g. Atlas : ). This follows the tradition cited by Prior (: ) that “all forms of negation are reducible to a suitably placed ‘It is not the case that.’ ” In some prior literature it is unclear whether the terms internal and external negation are so named because

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



    

The remainder of this chapter outlines how different approaches to modeling presupposition handle cases of cancellation while still accounting for their projection behavior. We will structure our discussion around four focal questions for which the major extant accounts of the interactions between presupposition and negation give different answers: Question I. Does the theory account for default projection through negation? Question II. Is default projection over negation the result of compositional semantics or pragmatics? Question III. Do the denial of entailments and the denial of presuppositions require separate negations? Question IV. Do the denial of conversational implicatures and the denial of presuppositions require separate negations? The rest of this section will give the reader a foundation for some of the main issues that arise for theories of presupposition and negation. Section .. describes Russell’s scope ambiguity approach to definite descriptions, an important precursor to modern theories of presupposition projection. Section .. comments on cross-linguistic work on presupposition and how these studies could inform our understanding of projection and cancellation. Section ., which follows a rough chronological organization, outlines a few of the major approaches to presupposition and negation: trivalence (..), underspecification (..), metalinguistic negation (..), cancellation (..), and accommodation (..). Section . summarizes the differences between these frameworks with respect to the four questions above and concludes.

... Presupposition projection as scope ambiguity Although Russell did not take there to be a separate class of presuppositional phenomena, his account of definite descriptions set some important groundwork for truth-functional approaches to presupposition projection. A brief detour into this work will be helpful for understanding the major questions that arise for the theories of presupposition and negation that will be detailed in section .. Consider the infamously ambiguous sentence in (a), which is said to have both the preservative reading illustrated by (b) and an annihilative counterpart that might be represented by (c).

of their syntactic position, or because the syntactically external negation tends to produce a wide scope reading, and thus is ‘external’ at some level of Logical Form representation. It is notable that as regards external negation, the scope facts and the presupposition projection facts seem prima facie quite different. Consider the sentence ‘It’s not the case that every linguist realized that the solution was in Horn’s dissertation all along,’ in which an external negation is combined with a quantifier (‘every’) and a presupposition trigger (‘realized’). We find it hard to get a reading for this in which the universal outscopes the negation, i.e. where the sentence means roughly ‘Every linguist failed to realize . . . ’ On the other hand, we certainly do get a preservational reading, whereby someone uttering the sentence would be taken to be committed to the relevant solution having been in Horn’s dissertation. Given that sometimes presupposition projection occurs even when a syntactically embedded operator cannot take wide scope, it is clear that presupposition projection facts cannot be explained by independently motivated scoping mechanisms alone.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   ()



a. The King of France is not bald. b. The King of France is ‘non-bald.’ c. It is not appropriate to assert ‘the King of France is bald.’⁶

Since there is no King of France whose existence can truthfully be presupposed, the truthvalue of (a) is in question. Russell’s () proposal for dealing with this data involved including an existential quantifier over a variable that can be ascribed a property by a separate predicate in the LF. For example, the King of France is bald would be formalized as (a). To get the ambiguity between a preservative and annihilative reading of (a), we can manipulate the scope of the negative operator in the LF, as in (b) and (c) respectively. () a. 9x (Kx & 8y[Ky → y=x] & Bx) b. 9x (Kx & 8y[Ky → y=x] & ¬Bx) c. ¬9x (Kx & 8y[Ky → y=x] & Bx) Scope differences nicely motivate the ambiguity of cases like (a), since preservative and annihilative readings can be mapped to different LFs. However, we run into issues if we try to extend a purely Russellian theory beyond the presuppositions of definite descriptions. For example, say we want to account for the presuppositions of stop. Since there are two meaning components implied—David’s current smoking habits and his past smoking habits ()—we can manipulate the scope of negation between these two bits of information. We could narrow the scope to target only the at-issue implication (a), or only the notat-issue implication (b). But in a Russellian theory the negation should also be able to outscope both implications, as in (c). Here, it is either true that David didn’t smoke in the past or that he does smoke now, although these truth conditions seem hard to grasp and it is certainly not intuitively obvious that there is any such reading. ()

David stopped smoking. → David was a smoker and is a non-smoker.

()

David didn’t stop smoking. → a. David was a smoker and is a smoker (is not a non-smoker) b. David never smoked (was a non-smoker and is a non-smoker) c. It’s not the case that David both was a smoker and is a non-smoker (Either he didn’t smoke, or he does smoke)

Russell’s theory operates within a bivalent truth-conditional system, putting what others would label as presuppositions directly into the LF. On such a view, presupposition projection can be attributed to compositional semantics rather than by some sort of pragmatic process. As we will see in section ., this is a trait that aligns a Russellian ⁶ The reading here is the one that might have been represented in the literature using e.g. ‘It is not the case that . . . ’. However, here we follow e.g. Horn () in using the more pragmatically oriented representation ‘It is not appropriate to assert . . . ’. While this choice is not entirely neutral since ‘appropriateness’ might suggest a pragmatic rather than truth-functional analysis, we do not intend that the representation of the reading carries any theoretical predilection as regards whether the analysis should be semantic or pragmatic.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



    

account with some (‘semantic’) theories of presupposition but differentiates it from others. Another important point is that Russell’s scope ambiguities do not explain why presuppositions project through negation by default. If we can freely alternate between readings, why is the narrow scope, preservative reading more likely, while the wide scope, annihilative version is more marked? To account for such data, Russell’s account would need to be augmented with a substantive pragmatic account of scope preferences. Many of the theories we will discuss come from the other direction; they are primarily cast as pragmatic theories of preferences over readings and marginalize the role of compositional semantics to varying degrees. A further issue differentiating accounts of negation and presupposition is whether the same negation operator can be used to target implicature and presupposition, but here Russell’s account simply doesn’t compare, since it was never designed to account for inferences we would now term implicatures in the first place.

... Cross-linguistic studies Before outlining the major models of presupposition and negation, let’s address one final aspect of the relevant data that they must account for: how presupposition and negation interact cross-linguistically. The vast majority of presuppositional research has been limited to English. However, cross-linguistic comparisons have the potential to shed light on some important issues related to the nature of presuppositions, and therefore the nature of the mechanisms through which they can and can’t be negated. For example, if we’re committed to the idea that implicatures and presuppositions are represented in different systems, we might expect there to be distinct mechanisms through which they are negated in some languages. However, even among languages with an array of negation strategies, linguists have yet to identify a language where those morphosyntactic negation strategies are allocated among different types of implications in such a way as to separate annihilation of implicatures and presuppositions. More broadly, if we expect an ambiguity between semantic negation and pragmatic denial signaling negation, we might also predict that some language has distinct morphology for accomplishing these functions. Again, there had not been evidence that this is the case in any language (see e.g. Gazdar a: –; Horn : –; van der Sandt and Maier :  for discussion). Another issue that has been taken up in existing cross-linguistic research has to do with the degree to which presuppositions are conventionally attached to their triggers. If presuppositions are conventionally associated with triggers in the same way that lexical items are conventionally associated with semantic content, we might expect that, like other aspects of lexical meaning, they could be detached from those triggers and would vary cross-linguistically. However, if instead presuppositions follow from the meaning and use of particular expressions, we would predict that the distribution and behavior of presupposition triggers would be uniform across different languages (see e.g. Simons ; Levinson and Annamalai ; von Fintel and Matthewson ; Prince  for discussion). If cross-linguistic evidence is strong enough to suggest that presuppositions are not conventional, theories that account for presuppositions via conventional semantic stipulations would have to adapt. A few studies have found evidence that English triggers have similarly presuppositional counterparts cross-linguistically. This might be used as part of an argument that the

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



presuppositions are not conventionally associated with the triggers, but rather that there is some general principle allowing the presuppositions to be derived from the ‘ordinary’ meaning used to judge synonomy of the trigger expressions across languages. For example, Levinson and Annamalai () find analogous presupposition triggers in English and Tamil, which behave similarly with respect to several properties associated with presuppositions, namely contextual defeasibility, explicit cancellation, and Karttunen’s () filtering phenomena. Tonhauser et al. () also observes similar projection behavior in both English and Paraguayan Guaraní (Tupí-Guaraní) through a range of embedding constructions beyond just morpho-syntactic negation.⁷ If all these presuppositions were conventional, then one might also anticipate a degree of arbitrariness, whereby the presuppositions associated with lexicalization of a concept varied from one language to the next. However, there has been at least one case where a major difference has been observed between presuppositions in two languages. Von Fintel and Matthewson () specify that, beyond the belief that all languages have presuppositions, most semanticists, regardless of theoretical orientation, are inclined to believe the following: ()

All languages allow their speaker to express aspects of meaning which a. are not asserted, but somehow taken for granted, b. impose some constraints on when an utterance is felicitous, and c. project through certain entailment-cancelling operators.

While the previously mentioned studies support (a) and (c), the findings of Matthewson () indicate that (b) may not be true for St’át’imcets and other Salish languages. She observes that in this language speakers don’t voice any objection in the case of a presupposition failure. In other words, listeners won’t explicitly deny a presupposition that they find to be false. Matthewson argues that presuppositions in St’át’imcets are not Stalnakerian () in the sense that they place constraints on the common ground. Rather, St’át’imcets presuppositions follow Gauker’s () model, such that they only belong to the speaker’s set of assumptions,⁸ so hearers have no grounds to question them. Evidence for presuppositional universals could impact the way that we model presupposition and negation, especially in terms of the conventionality of presuppositions and their comparability to entailments. But, contrarily, cross-linguistic differences should lead to parameterization of theories of presupposition so as to account for variation across languages. For example, if St’át’imcets presuppositions really do only belong to the speaker’s set of assumptions, there would be different parameters guiding the circumstances

⁷ These embedding constructions are sometimes referred to as the family of sentences diagnostic, named so by Chierchia and McConnell-Ginet (1990). This methodology gives the cross-linguistic researcher some flexibility in case a particular embedding construction cannot be implemented in the language of interest. ⁸ Gauker () is certainly not the only theory on the market that explores the possibility of non-common ground presuppositions. Soames’s () utterance presupposition involves no shared assumption but rather something a speaker expects to be uncontroversial. Kempson () considers whether presuppositions have to be assumed to be true by the hearer. Prince () discusses a variety of it-cleft which presupposes information that the speaker takes to be fact but is not known to the hearer.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



    

under which a speaker could object to the appropriateness of an utterance, and therefore when a speaker could use a metalinguistic negation operator. Similarly, if Gricean augmentation is used to motivate when we expect presuppositions to appear, the fact that St’át’imcets presuppositions don’t need to be in the common ground might change the types of contexts that make them pragmatically preferred.

.. A    

..................................................................................................................................

... Trivalence Half a century after Russell’s () scope ambiguity proposal, Strawson () resurrected the King of France data, bringing renewed attention to presuppositions as a distinct category of backgrounded information. Strawson’s key observation⁹ was that sentences where an expression fails to refer, like The King of France is bald, seem to be neither true nor false. Instead, he notes that they make the listener feel “very squeamish” when presented out of context (). The difference between truth-conditional falsehood and presupposition failure is difficult to pin down technically, but the latter certainly involves some sort of dissonance between what the speaker has taken for granted and the listener’s background beliefs. One indication of this dissonance can be seen in the ‘hey, wait a minute’ test. The first response in () indicates that the listener notices a difference between their background beliefs and the speaker’s presupposition that there is a person who has solved Goldbach’s Conjecture. On the other hand, the second response seems odd when atissue information is being questioned. ()

A: The mathematician who proved Goldbach’s Conjecture is a woman. B: Hey, wait a minute. I had no idea that someone proved Goldbach’s Conjecture. B0 : #Hey, wait a minute. I had no idea that that was a woman. (von Fintel )

Strawson captures this squeamish non-truthiness intuition by positing that when a presupposition fails to refer, the sentence simply lacks a truth value. So, presuppositions are definedness conditions for an utterance to have meaning. They are treated as propositions that are truth-functionally related to the utterance. So, one sentence presupposes another iff whenever the first is true the second is true, and whenever the negation of the first is true the second is also true. Crucially this relationship cannot hold between two sentences if either is undefined, so presupposition does not exist in cases where an expression does not refer. Since there are two distinct effects here, Strawson proposes that there are two distinct operators: a preservative negation operator and a presupposition-cancelling, annihilative operator. The crucial difference between the two is that the latter can create a true sentence from an undefined one, like when a presupposition is directly denied.

⁹ Following intuitions from Frege ().

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   ()

a. Preservative Negation Φ t f ★

¬Φ f t ★



b. Annihilative Negation Φ t f ★

#Φ f t t

So, let’s take our smoking example. A typical trivalent theory would predict the following truth values: () Danny stopped smoking. True if Danny smoked and no longer does True or False if Danny smoked, undefined if Danny never smoked ()

Danny didn’t [preservative negation] stop smoking. True if Danny smoked and still does True or False if Danny smoked, undefined if Danny never smoked

() Danny didn’t [annihilative negation] stop smoking. True if Danny never smoked or Danny has continuously smoked Always true or false, never undefined Some critics of a trivalent solution to presupposition failure take issue with the implementation of a third, undefined truth value. For example, Turner () points out that presupposition failure does not induce truth-valuelessness in and of itself because there are cases where presupposition failure results in the assignment of a classical value. Strawson () concedes this point himself. For example, () is true even though the existence presupposition fails: ()

The moon wasn’t hidden by the clouds because there weren’t any.

Trivalent accounts do not necessarily require ambiguous negation like Strawson’s. Indeed, Burton-Roberts (b) advocates for a trivalent theory with a single negation. While he achieves this through the use of a pragmatic weakening operation, it is also possible to maintain a single negation in a trivalent account using additional semantic resources. This is the approach taken by Beaver and Krahmer (), who suggest that a trivalent semantics with a single preservative negation is tenable if combined with the Assertion operator (here: A) of Bochvar (). This operator can be thought of as mapping a proposition F to the proposition that F is assertable, where a proposition is assertable if it is neither false nor undefined. ()

Bochvar’s assertion operator

Φ t f ★

AΦ t f f

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



    

This operator can be used to convert trivalent propositions into bivalent propositions, or, equivalently, presuppositions into entailments. So if we took A to be a silent operator at LF, () would have a reading representable as A(Danny stopped smoking), which would be true in models where Danny used to smoke but doesn’t now and false in all other models, even those in which the underlying sentence was undefined. Now consider the negated variant (). In addition to an A-free reading, there would be a reading where A outscopes negation, and one where it is outscoped by negation. It is this latter reading ¬A(Danny stopped smoking) that is of interest: it lacks any presupposition, and has truth conditions just like those in (). Indeed, more generally, it is clear that for any F, we have the equivalence #F  ¬AF, so provided we can accept the appearance of assertion operators in the LF, there is no need to posit the existence of a separate annihilative negation. Such an approach seems to allow a theorist deeply opposed to postulating ambiguity of negation to have one’s trivalent cake and eat it, preserving a uniform approach to negation, and yet accounting for cancellation of presuppositions. At the same time, this fortuitous result might just be seen as technical trickery, casting doubt on whether the question of whether negation itself is ambiguous was the right question to begin with. After all, a theory with an A-operator is one in which there is a semantic ambiguity (like e.g. a Russellian approach), even if that ambiguity is not a lexical ambiguity of negation. We note here just one potential benefit of the A-operator approach, and one, related, potential drawback. The good news is that the approach correctly predicts that cancellation readings should occur with operators other than negation; the bad news is that the theory, without further constraints, would be in danger of over-generating presupposition-free readings, failing to explain why presuppositions tend to project through a wide variety of embeddings and are only canceled in special circumstances. The most prominent alternatives to trivalent accounts involve a family of pragmatic theories, which we use here broadly to describe any theory where presuppositions are not accounted for by conventional stipulations and are not derived through semantic composition. Instead, pragmatic theories derive presuppositions through some process of inferencing or strengthening. In this sense, the first pragmatic theory can be attributed to Stalnaker (, , ),¹⁰ however there have been a wide range of theories since then that use a variety of pragmatic strategies to derive presuppositions, several of which we will describe in the coming sections.

... Underspecification Around the time that pragmatic accounts came onto the scene, several linguists had already argued that presuppositions associated with a wide range of linguistic constructions deserve

¹⁰ On Stalnaker’s account presuppositions are joint assumptions of both speakers and hearers in the common ground. It is unclear whether Stalnaker’s account should be taken to be a pure underspecification theory in the sense of section 20.2.2, since although he analyzes e.g. factive verbs as not involving any conventional stipulation of presuppositions, his account seems to allow that at least some expressions could depend for their semantic definedness on properties of the world and discourse context. That is, it is conceivable that there are both semantic and pragmatic presuppositions.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



a special status distinct from assertions (e.g. Fillmore ; Kiparsky and Kiparsky ; Langendoen and Savin ; Zwicky ), extending earlier work in philosophy of language that had focused on a more limited set of constructions, notably definite descriptions. However, other accounts (Stalnaker ; Atlas ; Kempson ; Boër and Lycan ; Wilson b) maintained that despite the commonality of projective behavior across constructions, a semantic distinction between presuppositions and assertions was not necessary. For them, both annihilative and preservative readings involve one semantic negation operator in fixed, wide scope. We term these theories underspecification theories (e.g. Beaver ),¹¹ since they assume that negative sentences express a single meaning which is semantically weak but can be pragmatically strengthened to a more specific, stronger meaning. From this perspective, presupposition is a Gricean augmentation of an utterance. Let’s look at how underspecification theories might account for our smoking example. For the affirmative sentence, we still get two inferences: that Danny smoked in the past and that he no longer smokes (). A negation of the sentence yields the reading in (a) since negation is taken to be in fixed wide scope over a complex proposition involving the conjunction of the presupposition and the at-issue content. It follows from simple application of De Morgan’s laws that the sentence is true if either of these implications is false, as in (b), which in turn is equivalent to (c). However, a pragmatic strengthening process leads hearers to infer that the presuppositional content is true. So, the truth conditions restricted to models where the presupposition holds become exactly the same as if the negation had targeted only the at-issue content. ()

Danny stopped smoking → Danny was a smoker and is a non-smoker

()

Danny didn’t stop smoking → a. It’s not the case that [Danny was a smoker and is a non-smoker] b. Either he didn’t smoke, or it’s not the case that he is a non-smoker c. Either he didn’t smoke (previously), or he smokes (now) d. After strengthening: He smokes (now)

Different underspecification theories propose different pragmatic motivations for speakers to use presupposition triggers, which in turn motivates why preservative readings are preferred. For Kempson (), the motivation is informativeness. A speaker will choose a trigger like a definite description to generate a conversational implicature, allowing the hearer to assume that the use of the definite is anaphoric and doesn’t require extra explanation. In Kempson’s theory, annihilative negation is not a semantic operator but is rather “one of the uses to which negative sentences could be put” (: ). For Wilson (b), the choice of using a presupposition is contingent upon a weighing of different options as being more or less likely choices for a speaker in a given context. In this theory, the grounds for which we can deny a presupposition can range from issues of how strong, ¹¹ Horn () refers to them as monoguist, as opposed to ambiguist theories which, as the name suggests, propose a semantic ambiguity of negation. Note that Burton-Roberts (a, b) proposes a reversed monoguist trivalent theory: he takes negation to be unambiguously preservative, and relies on pragmatics not for strengthening of weak readings, but rather for weakening of strong, projective readings.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



    

weak, or misleading a sentence is. Atlas () argues that rather than being ambiguous, negated sentences with presupposition triggers are non-specific.¹² Presuppositions then arise so that negative sentences may be informative and can relate to things that were already discussed in the discourse. Projection generally follows from a preference for logically stronger interpretations because presupposition strengthens the logical form. A more recent underspecification approach in which negation is assumed to take wide scope over presuppositional material is that of Schlenker (), who implements a pragmatic view of preferred expression choice and information update¹³ to derive presupposition projection effects. The heart of Schlenker’s account involves a set of competing pragmatic principles, the first of which is Be Articulate, which implies a preference for presuppositions to be articulated separately. For example, being explicit by articulating the presupposition in (a) and (b) would produce (c) and (d) respectively. So why is it that presuppositions aren’t always made explicit as in (c) and (d)? This is where a second principle which outranks Be Articulate comes into play: Be Brief. If ‘Danny smoked previously’ is already in the common ground, Be Brief implies that it will not be articulated. ()

a. b. c. d.

Danny stopped smoking. Danny didn’t stop smoking. Danny smoked previously, and he stopped smoking. Danny smoked previously, and he didn’t stop smoking.

But what about annihilative readings? When the presupposition is explicitly denied, we’re unable to explicitly add the presupposition as a separate conjunct, like in (a), on pain of inconsistency. So why in this case do we not explicitly insert the presupposition below the negation as in (b), which is what we might expect to find if we are being maximally articulate but avoiding inconsistency? Once again, Be Brief steps in, here leading to a preference for (c). The reasoning here is that adding the extra material would not change the interpretation at all: there is no difference in the truth conditions of (b) and (c), and no danger of the hearer taking the unarticulated presupposition to be in the common ground, given that the opposite (i.e. that Danny never smoked) is explicitly asserted. ()

a. #Danny smoked previously, and Danny didn’t stop smoking because he never smoked in the first place. b. It is not the case that Danny smoked previously and stopped, because he never smoked in the first place. c. Danny didn’t stop smoking because he never smoked in the first place.

The development of underspecification approaches was part of a lasting shift towards rooting the interaction of negation and presupposition in pragmatics rather than

¹² Atlas borrows this term from Zwicky and Sadock () but equates it to Quine ()’s general and Lakoff ()’s vague. ¹³ The dynamic account of information update means that Schlenker’s account is in some respects like the dynamic semantic views to be discussed below, but the semantics he uses is classical and bivalent, and the dynamics of information update plays only a very limited role in his treatment of negation.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



semantics, although the processes used to derive the inferences in underspecification theories have evolved significantly. We see glimmers of underspecification theories in later approaches where presuppositions are by default given a similar status to other types of inferences and are later differentiated through some sort of pragmatic enrichment.

... Metalinguistic negation Rather than proposing a semantic ambiguity between preservative and annihilative functions, Horn (, ) argues for a pragmatic ambiguity between truth-functional descriptive negation and what he calls metalinguistic uses of negation,¹⁴ which he claims is the way that presuppositions are canceled. The examples in () illustrate some common instances of metalinguistic negation: ()

a. (So you [mijənɨj̆d] to solve the problem.) No I didn’t [mijənɨj̆] to solve the problem—I [mæˈnɨj̆d] to solve the problem. b. I didn’t manage to trap two mongeese—I managed to trap two mongooses. c. They weren’t ‘gentlemen’—they were dudes. d. Some men aren’t chauvinists—all men are chauvinists.

Here, the aspect of the sentence that is being denied is the phonology (a), morphology (b), stylistic level (c), or, in (d), a scalar implicature. Metalinguistic negation is a speech act of objection, which can also be used to object to presuppositions: ()

a. I don’t regret inviting him—he jolly well gate-crashed. b. The King of France is not bald—there is no King of France!

Employing metalinguistic negation is a way of signaling that an inference is inappropriate or unassertable in a given context. So, for our smoking example, Horn would predict that our non-negated sentence with a presupposition trigger entails that Danny is a nonsmoker and implies that Danny was a smoker in the past (). The ordinary descriptive/ preservative negation () only targets the entailment, while the metalinguistic operator () either takes wide scope over both implications (a) or says that it is not appropriate to make this assertion in the given context (b). ()

Danny stopped smoking a. entails Danny is a non-smoker b. implies Danny was a smoker

()

Danny didn’t [preservative negation] stop smoking a. entails Danny is a smoker b. implies Danny was a smoker

¹⁴ For more detail on metalinguistic negation, see Chapter  of this volume.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 ()

     It’s not the case that Danny stopped smoking [metalinguistic negation] a. It is not both [entailed that Danny is a non-smoker and implied Danny was a smoker] b. It is not appropriate to assert that ‘Danny stopped smoking’

It is notable that Horn’s strategy could be implemented in a way that is akin to the neoBochvarian strategy of adding an A-operator, obtaining annihilative readings by adding an operator that indicates assertability within the scope of a standard negation. The difference between the Bochvarian strategy and Horn’s derive from the fact that Bochvar’s operator was designed to deal with the effects of classical truth-valuelessness alone, and so, unlike Horn’s assertability operator, would not account for the way negation can target for example phonological form or implicatures. Van der Sandt () follows Horn in arguing that presuppositions are cancelled via pragmatic denial, which he considers to be equivalent to negative assertions. For him, denying a presupposition is equivalent to removing that information from the discourse record on the grounds that the speaker objects to it.¹⁵ However, unlike Horn, van der Sandt generalizes denial to standard denials of truth-conditional content, and rejects the idea that a standard truth-functional operator is involved in “unmarked cases” (: ). Thus, in his framework the speech acts of denial and assertion are completely independent from morpho-syntactic negation. There are several contemporary accounts that utilize a notion of metalinguistic negation while maintaining (contra Horn) that there is only one, truth-functional negation operator (Carston , , ; Noh , ; Moeschler , c; Blochowiak and Grisot ). For these theories, there is not a distinction between descriptive and metalinguistic uses of negation, but rather a distinction between descriptive and metalinguistic interpretations of a negated proposition in context. These interpretations are determined at the level of what is communicated, beyond the levels of what is implicated and what is said. Among these theories, there are different explanations for why particular contexts induce metalinguistic interpretations. For example, Carston argues that when there is metarepresentational or echoic material¹⁶ under the scope of the negation operator, the proposition is pragmatically enriched with a metalinguistic reading. For Moeschler, the difference between descriptive and metalinguistic interpretations depends on what information already exists in the common ground. We’ve seen that the distinction between metalinguistic and descriptive interpretations can be analyzed in terms of a lower level metalinguistic operator or in terms of contextual factors. Regardless of where we find this metalinguistic distinction, what these theories have in common is that presuppositions can often be targeted by the same mechanisms that target other types of material, like style or morphology.

¹⁵ This kind of updating of the discourse record is characteristic of dynamic frameworks, discussed in section ... ¹⁶ See e.g. Sperber and Wilson () for more on metarepresentations and echoic use.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



... Cancellation Like underspecification and metalinguistic approaches, cancellation theories (e.g. Gazdar a, b; Soames , ; van der Sandt , ) involve a pragmatic account of preservation and annihilation. Underspecification and metalinguistic theories treat presuppositions like implicatures, so their validity must be determined by whether general, presumably Gricean principles of conversation favor the projective or non-projective readings, typically without direct reference to any conventional specification of what presupposition is triggered. In contrast, cancellation theories treat presuppositions as conventionally specified in the lexicon; however, what is specified is not a presupposition as such, but a potential presupposition, a proposition which is added by default, but which simply evaporates if it conflicts with other entailments and implicatures in the utterance or in the overall context. Gazdar describes the process of adding only those presuppositions which are consistent with other material as an all-the-news-that-fits strategy. For example, consider the following example from Soames (): ()

The King of France is in hiding. Q: There is a King of France. P: There is at least one king (of something).

()

Either the King of France is in hiding or there is no King of France.

Here, Q is a potential presupposition of () that also entails P. So, since nothing in the context conflicts with or cancels these presuppositions (i.e. the news that P ‘fits’), both Q and P will be presupposed by (). However, the disjunction in () creates a context where the speaker is not committed to the truth of either proposition. So, the potential presupposition in Q and propositions entailed by it like P are cancelled. Since presuppositions are also entailed for unembedded triggers, cancellation theories predict that both implications in our smoking example will be entailed in (). So, when we negate the sentence in (), both implications are eligible for cancellation. However, since the presupposition does not conflict with other entailments, it should be accepted. ()

Danny stopped smoking ) Danny was a smoker and is a non-smoker

()

Danny didn’t stop smoking ) Either he didn’t smoke or he does smoke Conventional potential presupposition: he did smoke

Note that cancellation theories are similar to underspecification theories in that they treat presuppositions like entailments, but instead of using pragmatic strengthening justified by general conversational principles to generate presuppositions, potential presuppositions are always generated by default and compete with other entailments and implicatures to see whether they will remain. This requires an ordering of different aspects of meaning, such that presuppositions are more vulnerable to cancellation than ordinary entailments. The fact that cancellation theories assume that presuppositions are defeasible by implicatures and entailments is somewhat unexpected, since there is an intuition that

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



    

presuppositions are stronger than implicatures. There is also an intuition that presuppositions should be added first when updating commitments, since they involve information that is assumed to be true before the ordinary semantic content, hence presupposition. Van der Sandt’s (, ) version of cancellation theory takes this intuition into account by defining projected presuppositions as those that can be fronted without violating neoGricean conversational principles. For example, consider (): ()

a. b. c. d.

There are some tarts and if the king is angry then the knave stole the tarts. There are some tarts and if there is a knave, then the knave stole the tarts. There is a knave and if the king is angry, then the knave stole the tarts. There is a knave and if there is a knave, then the knave stole the tarts.

In (a–c), fronting the presupposition does not conflict with any Gricean principles, therefore the fronted presupposition is predicted to project. However, in (d) the fronted presupposition yields a redundant utterance, and therefore it is not predicted to project.

... Dynamic semantics and accommodation In recent years, approaches to modelling presupposition projection have increasingly turned to dynamic frameworks, which treat utterances not as propositions but as functions which update the discourse context. In dynamic pragmatic theories (e.g. Stalnaker , ), presuppositions arise as joint commitments of speakers and hearers that are added to the discourse context. In dynamic semantic theories (e.g. Karttunen ; Heim , ; Beaver , ; Zeevat ; van Eijck ; Chierchia ), presuppositions are admittance conditions on contexts, meaning that in order for an utterance to be felicitous, it must entail all of its presuppositions. So, if in the course of updating contexts the current state of the dialogue does not already entail a triggered presupposition, “the listener is entitled and expected to extend it as required” (Karttunen a: ). In other words, the hearer will (to use Lewis’s  terminology) accommodate a triggered presupposition if it is not already satisfied in the discourse so that an update is still possible. The most prominent accounts of accommodation are found in Heim’s (, ) work and its dynamic semantic descendants, and in a separate DRT-based tradition initiated by van der Sandt () and developed especially by Geurts (b). We focus here on Heim’s model, where updating contexts involves adjusting the set of possible worlds that are compatible with an utterance. Assume that the context set of possible worlds is C and S is a sentence. Heim’s system includes rules like the following: ()

a. For a simple clause S, C + S = the subset of worlds in C that are compatible with S, but this is defined iff S’s presuppositions (if any) are true in all worlds in C. b. C + ¬S = C  (C + S)

This works fine if the presuppositions of S are satisfied in C, but what if they aren’t? We can think of Heim’s model of accommodation as involving an alternative method of update designed specially for this case, which adjusts contexts as needed in order to deal with failed presuppositions. Sometimes, when C + S isn’t defined because a presupposition of S fails to

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



hold in C, we can simply set C + S equal to C + P + S instead. For example, suppose in world CHANGE Mary smoked previously, but now doesn’t, in world NEVER she never smoked, and in world ALWAYS she smoked previously and still does. Let C = {CHANGE, NEVER, ALWAYS}. Let S be ‘Mary stopped smoking’. Then C + S isn’t defined because S’s presuppositions aren’t true in all worlds. However, C + P, that is {CHANGE, ALWAYS} is a context in which update with S is defined, so we calculate C + S as (C + P) + S, that is {CHANGE, ALWAYS} + S, that is {CHANGE}. How about when the presuppositions of a negative sentence fail? In that case, this strategy leads to two alternatives. Simply recognizing that the update C + ¬S is not defined immediately suggests that we might instead calculate (C + P) + ¬S. In the case at hand, this would mean updating {CHANGE, ALWAYS} with ¬S, producing {CHANGE, ALWAYS} – ({CHANGE, ALWAYS} + S), that is {CHANGE, ALWAYS} – {CHANGE}, that is {ALWAYS}. This approach, where the initial context is updated with the presupposition before the negative sentence is added, is what is called Global Accommodation. But there’s another way to make things work. Intuitively, the reason C + ¬S is not defined is that it contains an instance of C + S, which itself is not defined. How about we repair things there? In that case, instead of calculating C – (C + S), we will calculate C – (C + P + S). In the case at hand, we already calculated (C + P + S), and found it was {CHANGE}. So, C – (C + P + S) = {CHANGE, NEVER, ALWAYS} – {CHANGE} = {NEVER, ALWAYS}, that is the set of worlds where she either never smoked or smoked previously and still does. In other words, we have understood ‘Mary didn’t stop smoking’ as ‘It’s not the case that (she smoked previously and stopped smoking)’, which can be true if she never smoked, and can also be true if she did previously smoke and didn’t stop. This reading, where in effect the presupposition is interpreted as if in the scope of the negation, is precisely analogous to what would be achieved using wide scope negation in a Russellian theory, or using a wide scope annihilative negation in a trivalent account. This is what is termed the Local Accommodation reading, and, notably, is achieved without recourse to an ambiguity of negation. We can represent the approach, where update with a complex sentence is no longer fully determinate but rather involves multiple possibilities depending on which accommodation strategy is used, as follows: () a. C + ¬S = C  (C [+P] + S) [Local accommodation] OR = C [+P] + ¬S [Global accommodation] = C [+P] – (C [+P] + S) Heim () argues that speakers will only utilize local accommodation when they cannot accommodate presuppositions globally. This preference for global accommodation translates to a dispreference for cancellation, explaining why presuppositions will project by default. Once again, annihilative readings are achieved without recourse to an ambiguity of negation, though whether the analysis should be thought of as involving any semantic ambiguity at all depends on whether local accommodation should be thought of as semantic or pragmatic, an issue that is perhaps somewhat moot in the brave new world of dynamic semantics, where aspects of meaning traditionally thought of as pragmatic become conventionalized within the semantics. Whether the treatment of cancellation can

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



    

be thought of as pragmatic or not, it is worth noting that dynamic semantics, and local accommodation more specifically, were not designed to account for a full range of ‘pragmatic’ uses of negation, such as discussed in section .., above, so that this type of theory, like the multivalent and underspecification approaches above, would presumably need to be augmented with some account of metalinguistic negation in order to deal with, for example, negations used to mark phonological infelicity.

.. C

.................................................................................................................................. Levinson and Annamalai (: ) call presupposition “a thorn in the side of a theory of meaning.” The unique behavior that presuppositions display under negation, namely contextually dependent projection and cancellation, are the prickly point on that thorn. The ways presuppositions interact with negation has been intrinsic to defining the phenomenon of presupposition itself. So, as accounts of presupposition have changed, so have proposals regarding the ways negation interacts with them. For example, the shift from semantic to pragmatic notions of presupposition was paralleled by a shift in attention away from truth-functional operators and towards pragmatic denials. After over a century of work on the topic, research on presupposition has expanded far beyond the projection problem as originally conceived, for example as regards variation of projectivity across different presupposition triggers (e.g. Chierchia and McConnell-Ginet ; Simons ; Abusch , ; Abbott ; Klinedinst ; Abrusán ; Romoli , ; Simons et al. ; Tonhauser et al. ). Such data will lead to new desiderata for future theories of how presupposition and negation interact, but our intention has been to focus on a narrower core of data that has motivated the bulk of contemporary theories, which we have explored in terms of four basic questions. We now recap how the major classes of theory answer those questions. A first difference between the accounts we have considered is how they explain default preservation (Question I, above), i.e. why presuppositions tend to project through negation, as well as other operators, in the absence of contextual or linguistic material that indicates that the speaker does not accept the presupposition. The explanation of default projection provides perhaps the most pivotal issue separating different types of accounts: • Underspecification theories assume that presuppositions will project unless there are pragmatic grounds for canceling them, although the explanations for projective behavior vary widely. • Metalinguistic theories treat annihilation of presuppositions as involving a special, marked type of negation, with the preference for preservation occurring because the ordinary meaning of negation is, by stipulation, unmarked. • Cancellation theories account for default projection by adopting Gazdar’s all-thenews-that-fits strategy. • Accommodation theories account for preservation through a combination of preferring contexts in which update is possible, and preferring global to local accommodation when update is otherwise impossible.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



Most theorists concede that presuppositions are not exclusively semantic nor pragmatic entities and have relevant properties for both types of systems. For example, perhaps pragmatic conditions motivate speakers to background certain information, and this information comes to be conventionally associated with particular words and constructions. On the other hand, perhaps a conventional presupposition that is represented semantically can be canceled at the pragmatic level when it is no longer relevant. However, it should be clear that the theories outlined here differ as regards Question II, that is whether they see the projection properties of presuppositions as following more from how the compositional semantics are configured, as in trivalent approaches, or from principles of pragmatics, as in underspecification and cancellation models; accommodation in dynamic semantics and DRT arguably provides a sort of halfway house between these two types of view, where the possibility of update results from compositional (dynamic) semantics, but in which much of the work is done by accommodation, an extra-semantic process. Accounts also differ with respect to Question III, that is whether entailments and presuppositions require separate negations. For scope ambiguity theories, underspecification theories, cancellation theories, and accommodation theories, presuppositions and entailments are modeled similarly, so both types of inferences are subject to similar types of negation. Regarding Question IV, concerning the way that theories of presupposition explain denial of implicatures, as well as denial of phonological appropriacy and other metalinguistic properties of the utterance, there is somewhat less clarity. Obviously, a metalinguistic operator is designed to handle all such cases uniformly. However, metalinguistic

Table . Summary of presuppositional accounts of negated sentences Question:

Framework: Scope ambiguity

I. Does the theory account for default projection through negation?

II. Is projection the result of compositional semantics or pragmatics?

III. Do denial of entailments and denial of presuppositions require separate negations?

IV. Do denial of conversational implicatures and denial of presuppositions require separate negations?

no

semantics (+ pragmatic disambiguation)

no

yes

Trivalent ambiguity no

semantics

yes

yes

Underspecification

yes

pragmatics

no

yes

Metalinguistic negation

yes

pragmatics

yes

no

Cancellation

yes

pragmatics

no

no

Accommodation

yes

both

no

yes

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



    

approaches tend to be stated in a somewhat impressionistic way, making it hard to determine the detailed predictions of the theory as regards presupposition projection. In contrast, many semantic theories are stated in a technically precise way but are more limited in their range of application. For a trivalent theory, or for that matter for a neoRussellian theory, an underspecification theory or a dynamic semantic account, the denial of implicatures would presumably have to be accounted for by some separate pragmatic system, but this is generally not spelled out.¹⁷ In Gazdarian cancellation theories, by contrast, the mechanism which defeats implicatures is explicitly the same as that which defeats presuppositions, both being defined as potential implications, again with the allthe-news-that-fits principle, that is: if you can’t add it, don’t. However, it is notable that while cancellation accounts do provide a uniform treatment of negation as regards to denial of entailments, presuppositions, and implicatures, such accounts do not provide an analysis of a full range of uses of negation, for example denial of the appropriacy of phonological form. This completes our discussion of the main themes in prior work on the interaction between presupposition and negation; the similarities and differences between accounts are summarized in Table ..

¹⁷ Some contemporary accounts of implicature take at least a subset of implicatures to be generated via a compositional mechanism that allows for intrusion of pragmatic inferences, i.e. addition of implicatures within the LF (Chierchia ). In such accounts a subset of implicatures, perhaps the set of scalar implicatures, and perhaps a larger set of generalized implicatures, could be outscoped by a standard truth-functional negation, hence deriving the effects of implicature cancellation.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  ........................................................................................................................

NEGATIVE DEPENDENCIES ........................................................................................................................

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  ......................................................................................................................

   ......................................................................................................................

 . 

.. I

.................................................................................................................................. I this chapter we review a number of key issues that have been discussed in the literature on negative polarity. The term ‘polarity sensitivity’, first used in print by Baker (), has been introduced to refer to forms whose distribution was observed to polarize in negative or positive contexts, as indicated by the contrast about the acceptability of anything in (), and the lack of inverse scope reading for (a). ()

a. Louise did not buy anything. b. *Louise bought anything.

[*9¬, ¬9]

The data from the German item jemals (ever) in () provide confirmation that the marginality is due to the requirement of the polarity sensitive item and not to structural requirements, for instance a mere word category mismatch. The item jemals must occur in a context in which the proper semantic/pragmatic properties are accessible. The contrast in the pair in (a)–(b) shows that the construction becomes unacceptable when the context does not provide an (accessible) negation. On the contrary, the sentences in the pair are equally acceptable in (c)–(d), where jemals is replaced by gestern (yesterday), an item that is also an adverb but not a polarity sensitive one. ()

a. Kein Mann war jemals glücklich. man was ever happy no ‘No man was ever happy.’ b. *Ein Mann war jemals glücklich. a man was ever happy ‘A man was ever happy.’

German (Drenhaus, Saddy, and Frisch )

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 .  c. Kein Mann war gestern glücklich. no man was yesterday happy ‘No man was happy yesterday.’ d. Ein Mann war gestern glücklich. a man was yesterday happy ‘A man was happy yesterday.’

Negative polarity sensitivity is a phenomenon that cuts across syntactic categories. Negative polarity items (NPIs) are, for example, indefinites that occur in the scope of negation and have a special form in many languages (Klima ; Ladusaw ; Haspelmath ), such as any in English (), adverbials, such as yet illustrated in (a), DPs functioning as minimizers, such as a drop in (b), and verbal idiomatic expressions such as lift a finger in (c) (see Chapter  in this volume on minimizers and other types of polarity items). ()

a. Louise hasn’t watched that movie yet. b. Louise did not drink a drop. c. Louise has not lifted a finger lately.

NPIs have sometimes been opposed to positive polarity sensitive items (Baker ; van der Wouden ; Szabolcsi ). Although this terminology suggests a mirror effect, the interpretation of the data that would support such a symmetry is not consensual by far. In the remainder of the chapter, we will first present licensors and environments that can host NPIs in section .. Then, section . is dedicated to reviewing several analyses that have been proposed in the literature and defining notions that they use, and will conclude the presentation by recalling some contributions to the debate on licensing conditions that have come from experimental data. We will also point out potential interactions between polarity sensitivity and other phenomena. Finally, the brief presentation of lexical items and other NPI expressions in section . will offer the opportunity to broaden the empirical base of the phenomenon of polarity sensitivity presented in this chapter, and to mention some of the main subtypes of NPI discussed in the literature.

.. NPI : L  

.................................................................................................................................. NPIs are more often characterized indirectly by studying their distribution than by tackling directly their lexical semantics. Following the standard terminology, the distribution of NPIs is analyzed in terms of licensing, that is all occurrences of NPIs in a clause are licit when properly licensed. As a consequence, sentential negation not in (a) is referred to as the licensor or the trigger of the NPIs.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



Different ways of understanding the licensing requirements have been offered in the literature, in terms of expressions that serve as licensors or alternatively in terms of environments in which they are licensed. Under the view that NPIs need a licensor, NPIs are often expected to be in the scope of an operator with the right syntactic and semantic properties. Typically, sentential negation is a suitable licensor. The presence of negation in a structure, though, is not always sufficient for NPIs to turn out to be acceptable in a sentence, and NPIs in subject position do not seem to be licensed by clausal negation, as indicated by the contrast in (). ()

a. Daniel did not look at anyone. b. *Anyone did not look at Daniel.

Syntactically, the scope relation is typically defined by requiring the licensor to occur in a higher c-commanding position, to the left of the NPI at surface structure. Adding a precedence condition to the dominance relation enforced via c-command is meant to capture the acceptability contrast between (a) and (), under the assumption that sentential negation outscopes the subject, whether or not it c-commands a lower copy of the subject at LF. ()

*Anybody didn’t buy a book.

Potential exception to this traditional view is the occurrence of NPIs inside an indefinite nominal that semantically takes narrow scope with respect to sentential negation, as in example () with a stage level predicate, from Uribe-Echevarria (). ()

A doctor who knew anything about acupuncture was not available.

Compared to the fairly large agreement on the syntactic role of c-command, the semantic contexts where NPIs are allowed are more numerous. Scholars have been looking for a weaker more inclusive notion of negativity that can be common to all the environments where NPIs are licit. Suitable environments have been subsumed under the terms affective (Klima ), downward entailing (Ladusaw ; Zwarts ; von Fintel ) and nonveridical (Zwarts ; Giannakidou ). Some examples of suitable environments are clausemate sentential negation, see (a) and (a), higher clause sentential negation, for example (b), where the role of neg-raising verbs might be crucial, a negative quantifier, for example (c), and constituent negation expressed by negative prepositions such as senza (without) in (d). ()

a. She hasn’t seen anybody. b. She doesn’t think she saw anybody. c. Niemand heeft ook maar iets gezien nobody has anything seen ‘Nobody saw anything.’

Dutch

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 .  d. Ha chiuso la porta senza alcuna spiegazione S/he closed the door without any explanation ‘S/he closed the door without any explanation.’

Italian

Sentential complements of adversative predicates such as doubt (a) and of factive predicates associated with a negative expectation such as be surprised (b) license NPIs with some variation across languages. Contrasting expectations also are detected under licensing degree modifiers such as too (c), that introduce what has been called predications of excess. ()

a. I doubt that he has done anything. b. She was surprised that anything so gross could have been said. c. He was too tired to understand anything.

NPIs are acceptable in the restrictor of universal quantifiers that are not strictly distributive (a), for example every but not each, of existential determiners like few (b), in the restrictor of superlatives (c), and under adverbs such as hardly (d). ()

a. Alle, som nogensinde har været i Tivoli, ville Danish Everyone who ever has been to Tivoli would kunne huske det. (Jensen ) can remember it ‘Everyone who ever has been to Tivoli would be able to remember it.’ b. Few people who had any interest in linguistics attended the meeting. c. This is the most stupid question any person has ever asked. d. They showed hardly any interest in linguistics.

In some cases, the co-presence of other specific lexical items enhances the acceptability, for example the NPI adverb ever¹ in (c) and else in a comparative sentence, cf. (). ()

He was more tired than anyone else.

Finally, let us just mention some more contexts about whose licensing power much has been written, namely the antecedent of conditionals (a), the immediate scope² of restrictive adverbs such as only (b), and questions (c). Questions are NPI licensors in their own right in some theories (see Ladusaw ; Kadmon and Landman ; Krifka ; Van Rooy , among others, and the debate about whether they are downward monotone).

¹ Ever and its corresponding forms are facilitators for NPIs as well as for N-words—following the terminology introduced in Laka ()—in negative concord languages, see the data on Romance languages in Corblin and Tovena (). ² The NPI is in the immediate scope of the negation operator in the semantic representation iff (i) it occurs in a proposition that is the entire scope of negation, and (ii) within this proposition there are no logical elements intervening between it and negation, as per Linebarger (). For languages that have NPIs of universal-like nature, such as Korean according to Kim and Sells (), a universal NPI takes immediate wide scope with respect to negation.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



Alternatively, they simply are contexts that introduce licensors by hosting silent operators in their syntax (e.g. Nicolae ). ()

a. If you need anything, let me know. b. Only Daniel has eaten any kale. c. Did she see anything?

The sentences in ()–() offer a glimpse of the variety of licensing contexts. This type of review prompts a warning that claims might have to be tweaked, because many of the examples discussed are from the English language, and things do not stay the same across items and languages. Indeed, not all NPIs have the same spectrum of licensing conditions, or in other words NPIs are not all of the same type. For instance, it has been proposed that there is a distinction between degrees of strength among the licensors (Zwarts ), and a corresponding classification into strong vs weak polarity sensitive items (van der Wouden ) has been put forth. We come back to this point below. The issue of the identification of the items that are polarity sensitive is also highly relevant. In many European languages, the potential space of negative polarity items is partly shared with N-words and the phenomenon that goes under the term of negative concord. This issue is discussed in Chapter  on negative quantifiers, and in Chapter  on negative concord. Moreover, it is possible that licensing contexts are not the exact same set within language families. For instance, sentential comparatives cannot host NPIs/N-words in all Romance languages (Corblin and Tovena ), and they may form a gradient with questions.

.. C  

.................................................................................................................................. The various proposals put forth in the literature have contributed important insights to the study of the phenomenon, although they might not have yet succeeded in providing full cross-linguistic empirical coverage. Approaches to negative polarity focusing on licensors have captured their licensing conditions using tools from syntax, semantics, and pragmatics, cf. Fauconnier (a); Ladusaw (); Linebarger (); Laka () among many others—see the review of three decades of research provided by Tovena (b). Conversely, the study of suitable environments for negative polarity has often resorted to pragmatic notions such as inferences and presuppositions (cf. Kadmon and Landman ; Krifka ; Chierchia ; among many others). Interestingly, several researchers would mix and match resources to cover the relevant data (cf. Linebarger ; Giannakidou ; among many others).

... Lines of analysis It was already mentioned above that NPIs are subject to certain structural restrictions. Syntactic studies have mainly focused on the definition of locality constraints, and have assumed negation, some covert operator, or some formal feature as licensors, depending on

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 . 

the framework in which they are developed. Locality is typically cast in terms of c-command between licensor and item, although the option of tackling semantic scope of licensors in terms of c-command has also raised some criticisms (cf. Hoeksema ). Locality conditions have also been used to motivate the distinction between types of NPIs based on whether NPIs require to be licensed locally or not. Strict NPIs are expressions that require local licensors, such as a living soul, verbal idioms such as breathe a word, and possibly until, for example. These are contrasted with non-strict items such as any. Let us consider first a proposal put forth within the Government and Binding paradigm, where NPI licensing is analyzed using the geometrical constraints of Binding Theory severed from referential considerations. Progovac (, ) explores the hypothesis that NPIs are subject to principle A of Binding Theory, whereas positive polarity items are pronominal and are subject to principle B. NPIs are assumed to be similar to reflexives insofar as they must be licensed, and they have to be bound in the local IP at S-structure. Sentential negation in INFL is a potential binder for NPIs. If locality is not met at S-structure, movement is used to recreate it at LF. Quantificational NPIs whose morphological properties do not prohibit raising can raise, and this makes it possible for them to be bound locally by a covert operator or a distant superordinate negation that would otherwise be considered to fall outside the governing category IP. Next, in the Minimalist Program, licensing is defined as feature sensitivity, and it is dealt with in terms of formal feature checking. For example, Jensen () proposes that Danish NPIs such as overhovedet (at all) host a formal feature +PI (for Polarity Item) that is strong and must be checked at NegP (via movement), usually by Spellout. Limitations on interpretation of some Danish NPIs like these, whose licensing possibilities are restricted to negation, are captured by requiring a locality constraint. The option explored by Jensen is to posit a ΣP—borrowed from (Laka )—in every clause immediately dominated by IP, and to assume that ΣP subsumes NegP. Polarity sensitive items must check their feature at the closest ΣP, and only when negation is present is NegP activated. Within semantically based approaches to polarity sensitivity, the most influential is the theory put forward in Ladusaw (, a).³ This analysis proposes the same treatment for all the different polarity items, and aims at reducing all licensing contexts to a single notion, namely downward entailment. A formal definition of this property is as follows: A function f is downward entailing iff for all x,y such that x → y, f(y) → f(x) (Ladusaw ). Downward entailment, a.k.a. implication reversal or monotone decrease (cf. Fauconnier a) can be understood in terms of logical inferencing from sets to subsets, from the general to the specific. For instance, consider the entailment in (). ()

Daniel is a poultry farmer ) Daniel is a farmer

The validity of the entailment in () shows us that having the more specific property of being a poultry farmer entails having the more general property of being a farmer. This is

³ The boolean algebraic core of this proposal was explored by Zwarts () in his contemporary research. The relevance of logical entailments for polarity sensitivity was already pointed out by Baker () and discussed by Fauconnier (a).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



normally the case, but negative contexts reverse the direction of implication, as shown us by (). ()

Nobody is a farmer in our village ) Nobody is a poultry farmer in our village

The notion of downward entailment matches the empirical observation that NPIs typically contribute to strengthening negative contexts (Van Rooy ). The prediction of this theory is that NPIs are licensed in any environment or by any operator that is downward entailing. Negative operators do not exhaust NPI behavior, and it is worth underscoring that downward entailment covers sentential negation and other weaker forms of negative contexts, such as the scope of an adverb like hardly or the restrictor of all. Moreover, downward entailment is an independently motivated notion that was studied within the theory of generalized quantification, and has turned out to be handy in the study of various phenomena. If downward entailment is seen to correspond to a form of weak negation, a possible strategy to tackle licensing variation is to posit that licensors have various degrees of strength. A hierarchy of licensors has been defined in terms of partial/total satisfaction of De Morgan’s Laws by Zwarts (, ).⁴ Negative expressions like few N are characterized as downward entailing functions, and are labeled as a weak form of negation.⁵ They are low in the hierarchy, and conversely, the NPIs that are licensed by monotone decreasing functions are called weak, e.g. ooit (ever) and idiomatic een pretje (a bit of fun) in Dutch (van der Wouden ). Next, expressions like no N and negative elements like N-words express a stronger form of negation, as they express anti-additive functions. They can

⁴ De Morgan’s laws are a pair of transformation rules that allow the expression of conjunctions and disjunctions purely in terms of each other via negation. They can be expressed as the two biconditionals (P ∧ Q) $ ¬(¬P ∨ ¬Q) and (P ∨ Q) $ ¬(¬P ∧ ¬Q). Separating each biconditional, and generalizing from the complement function ¬ to arbitrary functions f, yields four implications: (i) (ii) (iii) (iv)

(P ∧ Q) → ¬(¬P ∨ ¬Q) (P ∧ Q) ¬(¬P ∨ ¬Q) (P ∨ Q) → ¬(¬P ∧ ¬Q) (P ∨ Q) ¬(¬P ∧ ¬Q)

Zwarts () notes that for monotone decreasing functions only (ii) and (iii) are valid. Anti-morphic expressions correspond to the whole of De Morgan’s laws. Anti-additivity turns out to correspond to (ii), (iii), and (iv). Finally, anti-multiplicativity implies the validity of (i), (ii), and (iii) (Zwarts ). This yields an implicational hierarchy where downward entailing functions include anti-additive functions that, in turn, include anti-morphic ones—anti-additive functions and anti-multiplicative functions being unordered between them. The latter functions correspond to non-lexical expressions such as negated universals, and are not always mentioned in general presentations of the hierarchy. A formal definition of the functions in set theory is as follows: (i) A function f is anti-additive iff for all x, y: f(x ∨ y) = f(x) ∧ f(y). (Zwarts ) (ii) A function f is anti-multiplicative iff f(x ∧ y)=f(x) ∨ f(y). (Zwarts ) A function f is anti-morphic iff f is anti-additive and anti-multiplicative. ⁵ An even weaker class of licensors has been identified via the property of non-veridicality, see below. This property is more inclusive and extends Zwarts’s hierarchy. On the contrary, anti-veridical contexts directly find their place in the hierarchy as strong licensors, given that they are anti-additive.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 . 

license strong NPIs, and conversely, NPIs are called strong if they require licensors that are anti-additive functions. For examples, the Dutch expression ook maar iets (anything) and verbal idioms such as met een vinger aanroeren (touch with a finger) are not licensed by nauwelijks (hardly) and weinig (few), and need niemand (nobody) or nooit (never), as illustrated in () from van der Wouden (). () a. *Nauwelijks heeft ook maar iemand een Hardly has any body a b. Nooit heeft deze student ook maar iets Never has this student any thing ‘This student has never achieved anything.’

woord gezegd. word said. gepresteerd. achieved.

Dutch

Finally, expressions such as it is not the case that correspond to anti-morphic functions, are considered to represent a very strong type of negation, and are governed by the two laws of De Morgan. They license super strong NPIs. Thus, NPIs are called super strong if they are licensed just by anti-additive functions. For example, Dutch idiomatic expression voor de poes (literally ‘for the cat’) is fine in anti-morphic contexts, such as niet voor de poes zijn (to not be considered lightly), and ungrammatical in anti-additive ones, as illustrated by the contrast in () (van der Wouden ). () a. *Die AIO is nooit voor de poes. That graduate student is never for the cat.

Dutch

b. Die AIO is geenszins voor de poes. That graduate student is not-at-all for the cat. ‘That graduate student is certainly not to be trifled with.’ As we see, the classification of NPIs is based on the single semantic notion of entailment, according to this line of analysis. The distinction between weak, strong, and superstrong negative polarity items corresponds to the characterization of licensors as expressions that are monotone decreasing, anti-additive, and anti-morphic functions (Zwarts ). In recent work (Gajewski ), the weak/strong opposition is replaced by a dichotomy based on a generalized notion of downward entailment, called Strawson Downward Entailment, that takes presuppositions into account.⁶ The relevance of this notion can be best appreciated when considering the licensing behavior of only. ()

Only John eats vegetables John eats kale ∴Only John eats kale

For Only John to be downward entailing and license NPIs in its scope, the first premise should entail the conclusion. However, we need to add the premise that ⁶ A formal definition of Strawson Downward Entailment is as follows: (i) A function f is Strawson downward entailing iff for all x, y such that x → y, and f(x) and f(y) are defined, f(y) → f(x). (von Fintel )

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



the conclusion’s presuppositions are satisfied for this to be reasonable. Then, if it is contextually true that John eats kale, Only John eats vegetables Strawson-entails Only John eats kale. Semantic theories based on downward entailment have become further more articulate with the introduction of the notion of (non-)veridicality proposed by Zwarts () and Giannakidou (). Non-veridicality offers a characterization more apt to capture polarity sensitivity across languages. Consider that propositional operators are proposition embedding functions. A propositional operator F is veridical if and only if whenever Fp is true, p is also true. If this does not hold, then F is non-veridical. Thus, the truth of a proposition cannot be guaranteed if the proposition appears within a non-veridical context. For instance, the truth of the proposition Daniel is at home is not guaranteed in the context Luise hopes__. Similarly, a dyadic connective such as and, regarded as a two place sentential operator, is veridical with respect to both argument positions, whereas or is non-veridical with respect to both arguments. Additionally, a non-veridical operator is anti-veridical if and only if whenever Fp is true, p is not true. All downward entailing contexts are non-veridical, and the characterization as non-veridical is intended to also capture the licensing power of modal verbs, cf. () from Giannakidou (), and questions (). This is a welcome result, considering that it has proved particularly difficult to pin down what makes questions to be environments that license NPIs (see Borkin ; Van Rooy ). Anti-veridical operators are operators such as negation and without, which are the prototypical licensors of NPIs. () O Janis bori na milisi me {kanenan/opjondipote} with anybody the John may  talk ‘John may talk to anybody.’

Greek

() Heb je ook maar iets gezien? anything seen have you ‘Did you see anything?’

Dutch

Forms of pragmatic enrichment have been proposed to remedy imprecise prediction of analyses based on downward entailment conditions, both for cases where the conditions are under-restrictive and over-restrictive (see e.g. Schmerling ; Heim ; Yoshimura ). For example, the entailment in (a) is underwritten by meaning inclusion, that is extensionally brussels sprout is an hyponym of green vegetable and its intensional characterization includes that for green vegetable. However, the inferential schema associated with the entailment in (a), where the second proposition is stronger than the first, does not apply mechanically. This is shown by the fact that (b) is acceptable, although it might be a matter of personal opinion, but (c) is felt to be awkward out of context, although its antecedent could be seen as more general than the antecedent in (b). The lesson to learn is that the acceptability conditions of NPIs also depend on the current common ground of the conversation, that is on the set of mutually believed propositions, as per Stalnaker (, ), as well as on presupposition satisfaction as pointed out by Fintel (), as discussed above.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 ()

 .  a. Nobody ate a green vegetable ) Nobody ate brussels sprouts. b. If you read the New York Times, you remain quite ignorant. c. # If you read any journal, you remain quite ignorant.

Attention on strengthening effects had been drawn by the work of Kadmon and Landman () on English any, beyond the problem of polarity sensitivity. For instance, (b) is perceived to convey a statement that is less tolerant to exceptions than (a). ()

a. An owl hunts mice. b. Any owl hunts mice.

The recent production on scalar inferences and alternative semantics is sizeable, see among many others the theory of scalar implicatures proposed by Chierchia (, ). The idea of a two-component analysis of NPIs that distinguishes semantic structure from pragmatic principles is further explored by Krifka (, , ), and is to be understood as part of a broader work on focus and information structure. Krifka treats NPIs as a special case of the various constructions that introduce alternatives, like expressions in focus. These alternatives constitute a lattice where the NPI is the least element in the lattice denoting the property. Positive polarity items are a dual case, they constitute a lattice where the item is the greatest element. This proposal subsumes theories based on scales, where alternative propositions are totally ordered preserving certain inferences that are often pragmatically constrained. In the scalar-based theory put forth by Israel (, , ), the scalar nature of NPI makes them refer to minimal quantitative or informational values, be they amounts, degrees, or intensity. Further contribution to the debate on licensing conditions has come from experimental data. Let’s look at intrusive licensing first, that is illusions of grammaticality in patently ungrammatical structures that are associated with processing slow-downs. In their study on the influence of hierarchical constituency and linear order of the negator, based on negative polarity sentences in German such as in (), Drenhaus, Saddy, and Frisch () recorded differences in judgments about unacceptable cases, namely sentence (c) is more often wrongly accepted than (b). ()

a. Kein Mann, der einen Bart hatte, war jemals glücklich. man who a beard had was ever happy no ‘No man who had a beard was ever happy.’

German

b. *Ein Mann, der einen Bart hatte, war jemals glücklich. a man who a beard had was ever happy ‘A man who had a beard was ever happy.’ c. *Ein Mann, der keinen Bart hatte, war jemals glücklich. a man who no beard had was ever happy ‘A man who had no beard was ever happy.’ Drenhaus, Saddy, and Frisch () used acceptability speeded judgment tasks and event-related brain potentials (ERPs). The differences of judgment they report suggest that the simple existence of a potential licensor for an NPI may be sufficient to alter both the time course and efficiency of processing. In general, the results of their experiments

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



supported the hypothesis that unlicensed negative polarity items are unacceptable on both semantic and syntactic grounds, which confirm the relevance of linguistic studies on licensing. But comparing results about the incorrect conditions, they noticed that subjects accepted more often those structures in which the negative element precedes the NPI although it is not c-commanding it, than they did for structures without negation at all. This type of ‘fake’ or inaccessible licensor appeared to enhance the acceptability of the structure in the judgment data and weaken the effect that reflects semantic integration problems. Vasishth et al. () have subsequently proposed that this phenomenon of intrusion of ungrammatical retrieval candidate licensors is a case of resolution of a dependency between different parts of a sentence, dependency established by specific linguistic relations. In a nutshell, the effect is analyzed as an unusual instance of dependency resolution triggered by a kind of similarity-based interference, that does not by itself require a new definition of accessibility for NPI licensors. Second, experimental research has brought out a form of correlation in which syntactic concerns of locality and semantic concerns of licensors’ strength can be seen to meet. More precisely, the relative strength of the NPI seems to correlate with the degree of locality the licensor is required to have in order to make the appearance of the NPI completely acceptable. In their experimental study of German NPIs, Richter and Radó () observe a result that is rather puzzling for the theory of a hierarchy of licensors, namely that weak NPIs turn out to be more acceptable than strong NPIs in neg-raising environments. However, this result may well reflect a balancing out between locality and strength in the case of weak NPIs. The presence of a more powerful class of licensors (anti-additive) seems to be somewhat compensating a reduced locality in the position of the licensor. Analogously, the effect of a less powerful licensor is enhanced by a structurally closer position. On the other hand, strong NPIs appear to be less acceptable with a syntactically distant licensor than with a local one.

... Potential interactions between polarity sensitivity and other phenomena NPIs and their restricted distribution has been used as a test in the study of other phenomena. At the same time, examples instantiating other phenomena have been used as arguments in support of a characterization as (super) sensitive to negative polarity. In this section we briefly review two such cases, and leave the issues completely open. The first point of contact to be discussed concerns three domains, namely NPI studies, analyses of syntactic movement, and aspect. Data on so-called strong NPI until have been used to support a movement analysis of neg-raising and vice-versa; see Horn (), Gajewski (), Romoli (), and Collins and Postal () for a variety of positions. At the same time, until has sometimes been classified as a special strong negative polarity item that is lexically ambiguous and/or that, contrary to all other temporal expressions considered sensitive to polarity, is subject to further aspectual restrictions (Karttunen b; Linebarger ; Vlach ; de Swart ; Giannakidou ; Condoravdi ,

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 . 

inter alia; and a contrasting view in Hitzeman ; Tovena a). Typical data are provided in ()–(). ()

a. Daniel did not awake until the alarm went off. b. * Daniel awoke until the alarm went off.

()

a. Daniel did not sleep until the sun rose. b. Daniel slept until the sun rose.

The contrast in () is explained by considering until an NPI, and ruling out (b) for licensing failure. The lack of contrast in () is then accounted for by postulating two different words, a durative until and a punctual one, and by imposing aspectual extra requirements on the former. Durative until, being an NPI, requires a sentence describing a homogeneous eventuality. Sentence (a) is ruled in by assuming that negation is a stativizer operator, applies to the non-homogeneous event of ‘waking up’, and forms a negative state which is homogeneous. Next, a sentence such as () is used as support for an analysis of until as strict negative polarity item, that is requiring clausemate licensor, and of negation as a stativizer operator. ()

I didn’t think that Daniel would get here until tomorrow.

However, data such as () with the analysis of until as strict negative polarity item have been used in support of transformational analyses of neg-raising. Until is said to be licensed in () because the negative originates in the same clause as the NPI, and it is subsequently raised. A second point of contact of NPI studies and analyses of other phenomena that is worth mentioning concerns the idiomatic nature of many NPIs, and the rhetorical interpretations of questions. The licensing of strong idiomatic expressions such as (not) give a red cent to someone (=not to give any money to someone) has been brought to bear by Sadock () in his discussion of the issue of rhetorical interpretations for questions, see (). ()

a. Did Louise ever give a red cent to Emmaus? b. What metropolitan newspaper is worth beans? c. Who would lift a finger for you?

The effect of uttering a rhetorical question in a conversation is that of an assertion, according to Sadock (). A question such as (a) is rhetorical because it admits only a negative answer (see also Han ). In a conversation, it amounts to the assertion of a proposition of the reverse polarity, and such a proposition is assumed to be known to all the participants in the conversation. Indeed, affirmative wh-questions such as (b) can be easily recognized as being rhetorical. Assume for simplicity that, in wh-questions, interrogative sentence radicals denote open propositions and that they should be reconstructed as functions that map the missing piece of the proposition—corresponding to the wh-element—to the whole proposition. Then, when the wh- element is replaced by a value in the open propositional structure for (b), and the answer is produced, such an answer is expected to be a negative sentence so as to provide a licensing context for the NPI. Hence the answer is a foregone conclusion. Looking at the phenomenon from the viewpoint on

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



NPIs, questions that host idiomatic NPIs often are perceived as biased. To Horn (), Sedivy (), and Ladusaw () inter alia, the stance that the speaker expresses with questions such as (b,c) is clearly that of the corresponding negatives no one.

.. L    

.................................................................................................................................. As mentioned above, the class of words and expressions that typically are collected under the term of polarity sensitive items are a rather heterogeneous multitude. The brief presentation in this last section is meant to broaden the empirical base of the phenomenon of polarity sensitivity, and to offer the opportunity to recall some of the main subtypes of NPI discussed in the literature. As a necessary premise, we have to acknowledge again that the primary and maybe only common denominator of the various lexical items or constructions that have been treated as NPIs is their sensitivity to sentential negation. As a matter of fact, the effects of their polarity sensitivity and the distributional variation they exhibit are manifold, complex, and mingled. Reviewing some of their differences provides some help to better understand why one may have the impression that the theoretical analyses found in the literature have not yet attained explanatory adequacy. A vast amount of the literature on polarity items reports on the distributional patterns exhibited by any, one of the most common English NPIs. This expression has its own peculiarities, and at the same time it represents the more general group of indefinites. English any can be used in a variety of polarity contexts, as illustrated in sections . and ., but can also function as a free choice item, in which case it occurs in modal and generic contexts. In either interpretation, any is a referentially deficient expression (cf. Davison ; Kempson ; Israel ; Tovena and Jayez ; Jayez and Tovena ; among others). The issue of whether there are one or two uses of any, a free choice any and polarity-sensitive one, is still open (see Lakoff b; Ladusaw ; Carlson , ; Linebarger ; Horn and Lee ; Dayal ; among others). Given this broad distribution and the variety of readings, another hotly debated issue is whether a formalization as a universal item is more suited than an existential one. The latter analysis is more frequently defended, but see Dayal () for an opposing view. Comparative and diachronic considerations have been brought into the discussion. For instance, Hoeksema () argues in favor of assuming two separate analyses for any in consideration of the fact that indefinite NPIs do not automatically have this double interpretation, and that Dutch enig, a cognate of any, is a polarity item but lacks a free choice interpretation. In all cases, NPI any, its indefinite cognates, and non-interrogative occurrences of wh-words in languages like Mandarin Chinese, work like variables in the scope of negation and do not admit referential uses. A similar functioning is found in the class of minimizers, to which we turn next. Minimizers represent another of the best-known classes of NPI. We recall some of their peculiarities here, and refer the interested reader to Chapter  in this volume for a broader coverage. The term is from Bolinger () and refers to expressions like a thing, a drop, as

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 . 

well as idiomatic expressions such as a red cent, etc. The class stands out for its size, its broadness, as it spans from the domain of entities and amounts to the temporal and spatial domains, and its form, as it contains many colorful complex idiomatic expressions, not just lexical items and simple phrases. An apparent reason for the classification of minimizers as NPIs is the semantic effect of strict non-referentiality that distinguishes their use. When these items occur in positive contexts, if they do, they are interpreted literally and denote a minimal quantity. However, when they occur in negative contexts, they may be interpreted metaphorically, and the negation denotes the absence of a minimal quantity, and hence the presence of no quantity at all (Horn : ). This functioning as variables for negated minimal quantities brings them closer to NPI any. Indeed, minimizers are at the origin of the negative concord system of languages such as French, for example negation pas comes from a minimizer ‘step’ first used with verbs of movement and then generalized to all verbs, according to the unfolding of Jespersen’s Cycle (Horn ). A sort of mirror image of minimizers is represented by maximizers, for example temporal expressions such as in weeks, in ages. The literature on minimizers and maximizers is huge. Their functioning is related to scalarity effects, see Israel (, , ); Hoeksema and Rullmann (); among many others. Temporal perspective adverbs like yet and anymore have been proposed as candidate negative polarity items, as opposed to already, still that would be positive polarity items, see for instance Baker () and Linebarger (). There is a vast literature on these adverbs with regard to their logical and aspectual behavior, see Abraham (); Löbner (); Michaelis (, ); König (); van der Auwera (); Hoepelman and Rohrer () just to mention a few. The examples in () are taken to make the point that yet requires a negative context, and (a) is associated with an implicature that Daniel’s arrival is going to take place later than expected, if it does take place. The implicature is specific to yet and does not follow from a mere characterization as NPI. ()

a. Daniel is not yet here. b. *Daniel is yet here.

The pair in () illustrates the distribution of already as a positive polarity item. Sentence (b) is associated with a presupposition that Daniel’s arrival took place earlier than expected. Sentence (a), if interpretable, is understood as challenging the presupposition, and negation does not affect directly the temporal relation between event time and utterance time. ()

a. *Daniel hasn’t already arrived. b. Daniel has already arrived.

Opinions diverge on the issue, and on whether the same classification applies to languages where these perspective adverbs may be expressed by the same lexical form, cf. Italian ancora (Tovena ) and the examples in (),⁷ and more generally Romance languages.

⁷ Sentences (a) and (b) belong to standard Italian, while (c) is found in the variety spoken in Northern Italy.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

   () a. Luisa non è ancora qui Luisa not is ANCORA here ‘Luisa is not yet here.’ b. Luisa è ancora qui Luisa is ANCORA here ‘Luisa is still here.’ c. Ho visto ancora fare questo. have seen ANCORA do this ‘I have already seen this done.’

 Italian

On the other hand, Old English showed a similar cluster of uses for yet. The ‘still’, ‘again’, ‘already’ and scalar readings were all initially available and have subsequently been lost in an increasing specialization of use of the item (cf. König and Closs Traugott  and König ). The effects that the characterization as NPI attempts to capture is the distribution of readings rather than items, and the possibility vs. difficulty or impossibility to project or accommodate the different presuppositions they are associated with. Löbner (, ) is one of the main proponents of the idea of capturing the relations between the German elements schon (already) and noch (still), and their corresponding forms in other languages, in terms of duality. He draws an analogy with quantifiers on the medieval Aristotelian Square of Opposition. Duality always involves two negations, and two operators are dual if and only if the inner negation of one is equivalent to the outer negation of the other. Duality conditions are claimed to be met for schon and noch, and for nicht mehr (not any more) and nocht nicht (not yet). Next, let’s consider a group of NPIs that is more sparsely represented across languages. Semimodals in Germanic languages such as need in English, cf. (), and its counterpart hoeven (Dutch) and brauchen (German), have been frequently treated as NPIs (cf. Hoeksema ; van der Wouden ; De Haan ; Duffley and Larrivée ; Johannessen ; Levine ). This characterization seems to be primarily due to distributional facts that are not coupled with a special interpretive behavior. ()

John doesn’t think we need worry about his presence at the meeting.

In order to appreciate this issue, let’s summarize the main data on need. From the syntactic point of view, two distinct needs have been distinguished in the literature. One need selects a bare VP, and is an auxiliary, as shown by properties such as contraction, see () and ellipsis of to in front of an infinitival verbal form, compare (), and the other need that selects an infinitive, is a full verb and is exemplified in (). ()

We needn’t worry about his presence at the meeting.

()

You do not need to worry about that, at this point.

Auxiliary need has distributional properties of a negative polarity item, for example it occurs in negative sentences and in questions, see the unacceptability of ().

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 ()

 .  *We need worry about his presence at the meeting.

Moreover, NPI auxiliary need has morpho-syntactic special status, because it also has the properties of a modal, such as the lack or banning of any inflectional distinctions, the lack of a gerundive form, and its linear position with respect to other auxiliaries. From the point of view of polarity sensitivity, a peculiarity of need is that it can linearly precede the licensor, see () and (). ()

We need not worry about his presence at the meeting.

()

You need worry about (being defeated by) no one, at this point.

(Levine )

In sum, some form of interaction between negation, modality, and polarity looks like the key factor linking the modal properties of auxiliary need and its distributional properties of an NPI (Levine ). However, the data in () and () are remarkable, because inverse licensing is unusual among NPIs, despite the amount of variation observed among licensing conditions. The last candidate to the NPI status that we are going to review is represented by what could be termed a construction or a grammatical feature more appropriately than a lexical form or a phrase. It is the infinitive found in Romeyka, an endangered Greek variety spoken in Pontus Turkey, and discussed by Sitaridou (). In this language, the infinitive surfaces as a complement to a negated past tense modal, as illustrated by (), although neither Standard Modern Greek nor Northern Pontic Greek have an infinitive. () Utš eporesa mairepsini. not can.. cook. ‘I could not cook.’

Romeyka (Sitaridou )

Sitaridou proposes that the Romeyka infinitive is a new type of NPI, and argues that it is licensed by anti-veridicality, not by morphological negation. In her view, licensing by antiveridicality places the Romeyka infinitive on a par with Germanic semimodals like English need and German brauchen. This proposal would explain the unavailability of the Romeyka infinitive in other non-veridical contexts such as questions, non-veridical conditionals, present and imperfect tense negated modals. In her analysis, Sitaridou insists on parallelisms with Romance polarity subjunctives. Romeyka infinitive, by virtue of the fact that it is nonfinite, shares a T-to-C dependency with Romance subjunctives. Finally, such a dependency may have rendered the Romeyka infinitive diachronically more prone to developing a dependency with a negative operator.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ......................................................................................................................

         ......................................................................................................................

 

.. I

.................................................................................................................................. M (Pott ; Wagenaar ; Bolinger ; Horn ; among others) are minimal measure-denoting expressions such as English a word, a wink, an inch, Catalan una engruna ‘a crumb’, un pèl ‘a hair’, Spanish un alma ‘a soul’, (una) pizca de ‘(a) pinch of ’, Dutch ook maar een rode cent ‘even a red cent’ (Rullmann ), or Chinese yiju hua ‘a word’ (Shyu ) that denote low endpoints on a scale and, unlike regular Polarity Items (PIs), give rise to an even-reading (cf. Pott : ; Wagenaar ; Fauconnier a, b; Schmerling ; Horn , ; Linebarger ; Heim ; Abels ; Giannakidou , ). This has motivated analyses of minimizers as containing a tacit even particle (cf. Lahiri ; Lee and Horn ; Eckardt and Csipak ; Tubau b), which would explain why minimizers align with PIs in syntactic distribution, but not semantically. For example, in English both minimizers and PIs can occur in interrogatives, but the former give rise to rhetorical readings while the latter do not (Guerzoni ; Abels ). Given the potential use of minimizers as negation strengtheners, they have most often been characterized as negative polarity items (NPIs, see Chapter  in this volume). It is shown in this chapter, however, that this cannot be said to hold cross-linguistically. As observed by Suleymanova and Hoeksema (: ), minimizers have a non-literal or idiomatic meaning (cf. Tubau b), which may coexist with a literal meaning as regular DPs. Compare, for example, the interpretation of the DP a word in (a) (literal reading) with its idiomatic reading as a minimizer, (b–c). Only in (b–c) can a word be

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

considered a PI that activates alternatives along a scale (with the minimizer being at the low end). ()

a. Speaker A: Say a word. Speaker B: ‘Chocolate.’

English

b. Speaker A: Did she say a word about the incident? Speaker B: No, she didn’t say anything. c. She didn’t say a word at all. In Catalan, minimizers have to be obligatorily headed by the particle ni ‘not even’ for their interpretation to be an idiomatic (non-literal) one, while in Spanish, ni is optional (Vallduví ; cf. section .). Furthermore, Catalan minimizers can only occur in negative contexts, while the distribution for Spanish minimizers is wider. Interestingly, the wider distribution of Spanish minimizers reduces to negative contexts when ni is present. This supports the view that while minimizers may be considered a semantic class, they come with some particular syntactic properties that vary from language to language (Suleymanova and Hoeksema : ) (e.g. the aforementioned obligatory use of the particle ni ‘not even’ in Catalan; the optional co-occurrence of the numeral ene ‘one’ after the negative determiner geen ‘no’ in Dutch, cf. Suleymanova and Hoeksema ; among others).¹ Maximizers, by contrast, are expressions denoting large quantities or extents such as English all the time in the world, an eternity, Catalan ni per tot l’or del món lit. not even for all the gold of the world, or Spanish pesar una tonelada lit. weigh a ton, and although they have not received as much attention in the literature as minimizers, they have in common with the latter that, as observed by Israel (), they do not uniformly correspond to a particular type of PI either. Hence, the aim of the present chapter is to explore the nature of minimizers and maximizers as different kinds of PIs in English, Catalan, and Spanish, and identify connections between these lexical items and other elements in the polarity landscape such as negative quantifiers and Negative Concord Items (NCIs).² The chapter is organized as follows. In section ., some assumptions on the definition and classification of PIs that underlie the discussion in further sections are outlined. Section . explores the distribution of maximizers and minimizers in English, while section . studies the distribution of maximizers and minimizers in Catalan and Spanish. Section . addresses the nature of vulgar (taboo word) minimizers in the three languages under study and, finally, section . summarizes and concludes the chapter.

¹ As reported by Hoeksema (), the class of minimizers is very heterogeneous, including (i) DPs (e.g. English a word, a thing, a syllable, an inch); (ii) adverbial minimizers (e.g. English in the least, one bit); (iii) minimizing predicates (e.g. English sleep a wink, lift a finger, give a damn); and (iv) vulgar (taboo) items (e.g. English a fucking thing, shit). The same heterogeneity is observed in Spanish and Catalan. ² For diachronic approaches to minimizers see, among others, Meillet (), Jespersen (), Croft (), Hoeksema (, ), Eckardt (), Mosegaard-Hansen (), Willis, Lucas, and Breitbarth (), and Wallage ().

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



.. D   PI    

.................................................................................................................................. A PI is a linguistic expression that is sensitive to the polarity of the context it occurs in (Giannakidou ). PIs that usually occur in negative contexts have been traditionally referred to in the literature as NPIs (see Chapter  in this volume), although their distribution might not be limited to the scope of negation. This latter observation has motivated the distinction between weak, strong, and superstrong NPIs (Zwarts , , ; van der Wouden , ), which is directly linked to the notion of Downward Entailment (DE) (Fauconnier a, b, ; Ladusaw ). These three different kinds of PIs are defined according to the kind of operators that can license them: weak NPIs (e.g. any, ever) can be licensed by any DE operator, while strong and superstrong NPIs have further constraints with respect to the operators that can license them. Strong NPIs (e.g. in years, punctual until) are only licensed by a subset of DE operators, namely those that are anti-additive (e.g. no, never, not) (Zwarts ; van der Wouden ), and superstrong NPIs (e.g. one bit) by a subset of anti-additive operators, namely those that are antimorphic (e.g. sentential negation). Those PIs that repel negation, by contrast, have been referred to in the literature as positive polarity items (PPIs). PPIs (e.g. English some, (), Catalan força ‘much’, (), and Spanish también ‘too’, ()) are incompatible with negation (as shown in the (a) examples) but, as is the case for weak NPIs, they can occur in questions and conditionals (the (b) examples). ()

a. I would (*not) like some coffee. b. Would you like some coffee?

English

()

a. La Laia (*no) menja força. the Laia (*not) eats much ‘Laia eats quite a lot.’

Catalan

b. Si en Joan menja força, creixerà molt. if the Joan eats much grow.. much ‘If Joan eats a lot, he will grow up a lot.’ ()

a. María (*no) viene también. María (*not) comes too ‘María is coming too.’ b. ¿Viene María también? comes María too ‘Is María coming too?’

Spanish

In light of the heterogeneous licensing requirements of different kinds of PIs, Giannakidou () suggests (i) restricting the use of the label ‘NPI’ to PIs that are exclusively licensed by

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

negation, and (ii) embracing (non)veridicality—of which DE is a subproperty—as the core property of PI-licensing operators. A PI, therefore, would be defined as in (), with the different particular licensing conditions associated with different types of PI following from the various particular subproperties of (non)veridicality: ()

A linguistic expression α is a PI iff: (i) The distribution of α is limited by sensitivity to some semantic property β of the context of appearance; and (ii) β is (non-)veridicality, or a subproperty thereof: β ∈ {veridicality, nonveridicality, anti-veridicality, modality, intensionality, extensionality, episodicity, downward entailingness} (Giannakidou : )

Hoeksema () suggests a classification for different types of PIs that combines the strong/weak divide in Zwarts ( and ff.) and van der Wouden ( and ff.) with Giannakidou’s ( and ff.) approach to polarity and calls it ‘the extended Zwarts’ hierarchy’. As shown in Table ., the model features a concentric four-level classification where each type of PI is licensed by an increasingly restrictive subset of operators. In the present chapter, I explore how maximizers and minimizers fit into this hierarchy. I reserve the use of the term ‘NPI’ for strong and superstrong PIs, while I use ‘API’ (Affective PI, cf. Klima ) for weak and superweak PIs.

Table . Extended Zwarts’ hierarchy for PIs (Hoeksema ) Type of PI

Licensing operators Anti-morphic

Anti-additive

Downward-entailing

Non- veridical

Superweak

+

+

+

+

Weak

+

+

+

–

Strong

+

+

–

–

Superstrong

+

–

–

–

In addition, given that it has been suggested in the literature that PPIs (Baker ; Szabolcsi ; Nilsen ; Ernst ) are anti-licensed in the contexts where APIs and NPIs are licensed, they can also be modeled into the typology in Table . by looking at their potential anti-licensors (see Table .).³ Last but not least, while APIs and NPIs are not licensed in veridical contexts, all PPIs are felicitous in them. Returning to minimizers and maximizers, let us point out that Israel () observes that the former can be used to strengthen negation (i.e. they can be APIs, or NPIs, see

³ Notice that the parallelism that is being established requires that a type of superweak PPI be hypothesized, too. This is not without problems, though: these PPIs actually lack anti-licensors, which questions the role of anti-licensing as the defining property of PPIs.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



Table . Typology of PIs combining Hoeksema () and van der Wouden () Type of PI APIs NPIs

PPIs

Superweak Weak Strong Superstrong

Superweak Weak Strong Superstrong

Licensing operators Anti-morphic Anti-additive + + + + + + + – Anti-licensing operators Anti-morphic Anti-additive – – + – + + + +

DE + + – –

Non - veridical + – – –

DE – – – +

Non-veridical – – – –

Table .), but it is also possible to find them in the form of emphatic PPIs. Similarly, he notes that some maximizers are PPIs, while others are NPIs. Establishing what type of PI minimizers and maximizers are and exploring their particular (anti-)licensing requirements across languages is thus an interesting research avenue. In the remainder of the chapter, I consider how minimizers and maximizers in English (section .), and Catalan and Spanish (section .) fit into the classification given in Table ., and relate them to other elements in the constellation of polarity sensitive items such as negative quantifiers (English), and NCIs in Catalan and Spanish.

.. M    E

.................................................................................................................................. In English, a maximizer such as in weeks/months/years can occur in negative sentences that contain the negative marker (i.e. an anti-morphic operator), (a), and also an anti-additive operator such as nobody, (b). They cannot occur in the scope of a non-veridical DE operator, though, (c). According to the classification in Table ., therefore, the maximizer in weeks/months/years is a strong NPI. ()

a. She has *(not) visited me in weeks/months/years. b. Nobody has visited me in weeks/months/years. c. *Few students have visited me in weeks/months/years.

By contrast, other maximizers such as for all the tea in China, in a dog’s age, and in donkey’s years are only licensed by an anti-morphic operator, hence being superstrong NPIs, ().

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 ()

  a. I would *(not) do it for all the tea in China. b. *Nobody would do it for all the tea in China.

Variants of these expressions (namely all the tea in China and for donkey’s years), however, show a dramatically different distribution. They occur in veridical contexts, (a), in non-veridical—but not DE—contexts (e.g. under the scope of a modal, b), but not in DE contexts (e.g. under the scope of few, c). They are thus superstrong PPIs.⁴ ()

a. I *(don’t) love you all the tea in China. b. You could give me all the tea in China and I still wouldn’t live in a big city. c. *Few people love you all the tea in China.

Other maximizers such as all the time in the world, an eternity, for ages, or go to great lengths can occur under the scope of an anti-morphic operator, (a) and (a), an antiadditive operator, (b) and (b), a DE operator, (c) and (c), and a non-veridical operator, (d) and (d). As these expressions are also fine in a veridical context, (e) and (e), we conclude that they can be classified as superweak PPIs, the type of PPI we hypothesized in Table . when establishing a parallelism with APIs and NPIs. Interestingly, the Catalan and Spanish counterparts of these maximizers (see section .) show a similar behavior. Furthermore, as will be seen later in the section, no such type of PI is attested among minimizers. ()

()

a. b. c. d.

Hurry up! We don’t have all the time in the world! Nobody has all the time in the world to answer every email at once. Few scholars have all the time in the world to carry out research. If I had all the time in the world, I would read every single book in the British Library. e. There’s no need to rush. We have all the time in the world. a. b. c. d. e.

I don’t go to great lengths to check my email when I am on holiday. Nobody goes to great lengths to get an old computer repaired nowadays. Few people go to great lengths to inform themselves about their consumer rights. Do you go to great lengths to avoid confrontation? She went to great lengths to hide her feelings.

Turning now to minimizers, some are clearly superstrong NPIs (e.g. one bit), as they are only compatible with contexts containing an anti-morphic operator, (). ()

a. b. c. d.

I wasn’t one bit pleased with the results. Nobody is one bit pleased with the results. *Few students are one bit pleased with the results. *If she is one bit pleased with the results, she should say it.

⁴ A similar situation has been reported for French peu (little) and un peu (a little). According to Ducrot (), the former behaves as an NPI, while the latter behaves as a PPI.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



Others such as a word, have a clue, bat an eye, give a damn, or budge an inch are superweak APIs, as they are licensed by negation (both by anti-morphic and anti-additive operators, (a, b)), by DE operators such as few, (c), and in a variety of other non-veridical contexts such as questions (Ladusaw ; Haspelmath ; Giannakidou , ), (d), or conditionals (Haspelmath ; Giannakidou , ), (e).⁵ ()

a. b. c. d. e.

I didn’t say a word about it. Nobody had a clue about what to do next. Few batted an eye when their freedom of speech was threatened. Who gives a damn about the new company policy? If John had budged an inch, there wouldn’t have been a fight.

So far in this chapter, English minimizers have been characterized as APIs or as NPIs. Israel (), nonetheless, points out that some English minimizers such as of one’s own shadow and knock over with a feather behave as PPIs. A closer examination of these items in the light of the fine-grained classification laid out in Table . reveals that they should be considered weak PPIs. As shown in () and (), these expressions are anti-licensed by anti-morphic, anti-additive, and DE operators, (a–c) and (a–c), but are fine in nonveridical contexts that are not DE, (d) and (d). () a. b. c. d.

Godfrey is (*not) scared of his own shadow. *Nobody is scared of their own shadow. *Few people are scared of their own shadow these days. Are you scared of your own shadow?

(Israel b: )

⁵ Expressions such as lift a finger and sleep a wink present an extra complication: the grammaticality judgments for these expressions under the scope of a DE operator such as few are contradictory. For example, for Giannakidou (), Eckardt (), Eckardt and Csipak () these constructions are ungrammatical, while for Atlas () they are just deviant. Van Eijck (), by contrast, considers them grammatical. If they are ungrammatical, then they constitute a problem for the typology in Table ., as the model is concentric (i.e. if a PI is licensed in a given context of Table ., then it is predicted that this PI will be licensed in the rest of the contexts to the left). Hence the grammaticality of lift a finger and sleep a wink in non-veridical contexts predicts their grammaticality under the scope of few. It seems, though, that these expressions are in use (see the examples in (i), both from media material published in the UK). This is consistent with some other data such as (ii), which show that lift a finger is allowed in the scope of at most, also a DE operator (Pietarinen ). (i) a. Few lifted a finger to stop scab coal being brought in through the same ports to break the miners strike. () b. Then they brought in laws trying to curtail his/our right to protest outside Parliament and very few lifted a finger to do anything about that. () (ii) At most two people lifted a finger to help. (Pietarinen : )

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

Table . English maximizers and minimizers as different types of PI Type of PI

Maximizers

NPIs

Superstrong

for all the tea in China; in a dog’s age; in donkey’s years

Strong

in weeks/months/years

Superweak

all the time in the world; an eternity; go to great lengths

Superstrong

all the tea in China; for donkey’s years; donkey’s years ago

PPIs

Type of PI

Minimizers

APIs

Superweak

a word; have a clue; give a damn; bat an eye; budge an inch sleep a wink; lift a finger (but see fn. 4).

NPIs

Superstrong

one bit

PPIs

Weak

of one’s own shadow; knock someone over with a feather

()

a. b. c. d.

*She didn’t knock me over with a feather. *She never knocked me over with a feather. *No bodybuilder knocked me over with a feather. She could knock me over with a feather.

The observations drawn from the English data on maximizers and minimizers presented above are summarized in Table .. As can be seen, both classes of measuredenoting expressions contain items belonging to different types of PI described in section .. Let us now zoom into the class of minimizers that can be licensed by negation (i.e. APIs and NPIs), as their syntactic behavior has one interesting particularity. As discussed in Tubau (b), while API-minimizers clearly behave as PIs under the scope of sentential negation and under the scope of other non-veridical operators, they are ambiguous between a minimizer reading and a literal reading when negation is not sentential. Consider, for example, (), which, in the absence of a context that disambiguates the intended reading, can receive the two interpretations in (): ()

Mary said not a word.

()

a. Mary said nothing at all. b. Mary said not a word (but a full sentence). (Tubau b: , examples () and ())

While the reading in (a) corresponds to the interpretation of the minimizer a word as a strengthener of negation, a word is interpreted as an existential DP in (b). Interestingly,

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



as shown in (), the negation in (b) can be diagnosed as non-sentential.⁶ That is, if Klima’s () tests are applied to (b), a positive (reverse polarity) question tag is not possible, (a), neither-clause continuation is not possible, (b), either-licensing is not possible, (c), and not even-continuation is not possible, either, (d). If the tests are applied to a sentence such as (a), opposite results emerge, (a–d). ()

a. b. c. d.

Mary said not a word (but a full sentence), didn’t she? / *did she? Mary said not a word (but a full sentence) and so did Jane / *and neither did Jane. Mary said not a word (but a full sentence), too / *either. Mary said not a word (but a full sentence), *not even when asked to.

()

a. b. c. d.

I didn’t say a word about it, *didn’t I? / did I? I didn’t say a word about it, *and so did Jane / and neither did Jane. I didn’t say a word about it, *too / either. I didn’t say a word about it, not even after she insulted me.

It seems, therefore, that sentential negation and the idiomatic reading of the minimizer go hand in hand, and that the minimizer in () behaves similarly to a negative quantifier (see Chapter  in this volume). The quantifier-like behavior of minimizers that are adjacent to negation (e.g. ()) is further confirmed by a number of tests that have traditionally been used in the literature to examine the quantificational nature of polarity-sensitive items. According to Vallduví (), PIs fail tests (a, b), and pass (c), while this is the opposite for negative quantifiers.⁷ ()

a. Ability to occur in isolation. b. Grammaticality in preverbal position. c. Ability to appear in yes/no and if contexts with a non-negative value. (Adapted from Vallduví : )

By applying the tests in () to English (superweak) API-minimizers such as a soul, a word, lift a finger, and give a damn, Tubau (b: ) shows that when negation is adjacent to the minimizing expression, this successfully goes through tests (a, b), but fails (c), thus aligning with negative quantifiers (e.g. nobody, nothing) and not with PIs (e.g. anybody, anything). This is also the case for the (superstrong) NPI-minimizer one bit. By contrast, if sentential negation is not adjacent to the minimizer, the result is the reverse. This can be seen in examples ()–().

⁶ This kind of negation has been referred to in the literature as metalinguistic negation (Horn ) and it is generally assumed that it operates on a different level than descriptive negation. This might be the reason for (b) failing to successfully go through Klima’s tests for sentential (i.e. descriptive) negation. ⁷ Vallduví () also includes the ability to be modified by almost and absolutely. I have excluded it because there is some debate about what exactly it tests (cf. Giannakidou ; Horn a).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

()

Ability to occur in isolation Q: Who came to the party? a. A: Not a soul. /*A soul. b. A: Nobody. / *Anybody.

()

Grammaticality in preverbal position a. *(Not) a soul was waiting for John outside. b. Nobody /*Anybody was waiting for John outside.

()

Ability to appear in yes/no questions and if contexts with a non-negative value a. Did she lift (*not) a finger to help you? a0 . If you give (*not) a damn, call her tonight. b. Did *nobody / anybody help you?⁸ b0 . If you see *nobody / anybody waiting for me, tell them I am late.

Given that minimizers have been claimed to be associated with the Focus particle even (Linebarger ; Heim ; Horn ; Giannakidou ; Kuno ), which can be overt or covert, Tubau (b: ) assumes the structure of minimizers to be the one in ().⁹ ()

[FocP even [Foc] [NumP [Num a/one] [NP word]]]

As focused constituents can undergo Quantifier Raising (QR) (Rooth ), merging Negation to the minimizer (a Focus Phrase) allows the resulting NegP to QR to a left-peripheral position with wide scope (cf. Zeijlstra ), thus explaining the parallels between minimizers with adjacent negation and negative quantifiers presented in ()–().¹⁰ In the absence of negation, minimizers—with the syntax in ()—can be licensed in questions and conditionals, with the particle even supplying the emphatic meaning that is typical from minimizing expressions. Further research should determine whether minimizers also show similar parallels with negative quantifiers in those languages that have them.

.. M    C  S

.................................................................................................................................. In Catalan and Spanish, as was the case for English, maximizers can be classified as different types of PIs. For example, Catalan tot el temps del món ‘all the time in the

⁸ This question is felicitous with nobody if the speaker intends to enquire about the truth of the proposition nobody help you (i.e. is it true that nobody helped you?), but nobody does not have a nonnegative value. ⁹ As noted in Suleymanova and Hoeksema (: ), it is frequent to find the equivalent word for ‘even’ in minimizers across languages. This is ook maar in Dutch, and auch nur in German (Hoeksema and Rullmann ), bhii in Hindi (Vasishth ), sem in Hungarian (Surányi ), -to in Korean (Lee ), ere in Basque (Etxepare ). ¹⁰ See Haspelmath () for examples of indefinites formed with negative focus particles in different languages.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



world’ and its Spanish counterpart todo el tiempo del mundo qualify as superweak PPIs, as they can be licensed by anti-morphic, anti-additive, and DE operators, (b–d), as well as other non-veridical operators, (e). As was the case with certain English maximizers (cf. () and ()), the aforementioned expressions are also fine in a veridical context, (a). ()

món. a. Està jubilada i té tot el temps del is retired and has all the time of.the world ‘She is retired and has all the time in the world.’ b. Va, que no tinc tot el temps del món! come on that not have. all the time of.the world ‘Come on, I don’t have all the time in the world!’

Catalan

c. Nadie tiene todo el tiempo del mundo para arreglar sus asuntos. Spanish nobody has all the time of.the world to fix their problems ‘Nobody has all the time in the world to fix their problems.’ () d. Poca gente tiene todo el tiempo del mundo para escribir una tesis. few people has all the time of.the world to write a thesis ‘Few people have all the time in the world to write a thesis.’ e. Si tuviera todo el tiempo del mundo, aprendería una if had. all the time of.the world would.learn. one lengua tras otra. language after other ‘If I had all the time in the world, I would learn one language after another.’ Catalan trigar la vida / Spanish tardar la vida (y media) lit. last the life (and a half ) ‘last forever’, Catalan una eternitat / Spanish una eternidad ‘an eternity’, Spanish cantidad lit. quantity ‘a lot’, and Spanish pasarse tres pueblos lit. pass three towns ‘pass the limit’ are superstrong PPIs: they are anti-licensed in anti-morphic, anti-additive, and DE contexts, but are fine in non-veridical and veridical contexts, () and (). ()

a. (*No) trigaré la vida a acabar això! Catalan not last. . the life in finish this ‘It will take me forever to finish this!’ b. *Nadie tarda la vida en acabar un ejercicio. Spanish nobody lasts the life in finish an exercise c. *Pocs triguen la vida a acabar l’examen. Catalan few last. the life in finish the.exam d. Si tardan la vida y media en entregar el paquete, no pagues. Spanish if last. the life and half in deliver the parcel not pay.. ‘If it takes them ages to deliver the parcel, don’t pay.’

()

ha pasado tres pueblos con él, pobre. a. (*No) se not  has passed three towns with him poor ‘S/he has been too harsh on him, the poor guy.’ b. *Nadie se ha pasado tres pueblos contigo. nobody  has passed three towns with.you

Spanish

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



  c. *Poca gente se ha pasado tres pueblos contigo. few people  has passed three towns with.you d. ¿Se han pasado tres pueblos contigo?  have passed three towns with.you ‘Have they been too harsh on you?’

Spanish en días/meses/años/siglos lit. in days/months/years/centuries ‘in days/months/ years’ is a good example of a strong NPI. It is licensed by negation (both anti-morphic and anti-additive), but not by DE or non-veridical operators. ()

a. *(No) la he visto en días/ meses/ años/ siglos. not her have.. seen in days months years centuries ‘I have *(not) seen her in days/months/years.’ b. Nadie la ha visto en días/ meses/ años/ siglos. nobody her has seen in days months years centuries ‘Nobody has seen her in days/months/years.’

Spanish

c. *Poca gente la ha visto en días/ meses/ años/ siglos. few people her has seen in days months years centuries d. *Si la has visto en días/ meses/ años/ siglos, llámame. if her have. seen in days months years centuries call.me Finally maximizers headed by the particle ni ‘not even’ exist both in Catalan and in Spanish (e.g Catalan *(ni) per tot l’or del món / Spanish *(ni) por todo el oro del mundo lit. not even for all the gold of the world; Spanish *(ni) por lo más sagrado lit. not even for the most sacred). These expressions—obligatorily headed by the particle ni—can only be licensed by an anti-morphic operator, thus corresponding to superstrong NPIs in our classification. ()

a. *(No) vindria ni per tot l’or del món. not would.come not.even for all the.gold of.the world ‘I wouldn’t come for all the tea in China.’ b. *Nadie lo haria ni por todo el nobody it would.do not.even for all the c. *Ho faries ni per tot l’or it would.do. not.even for all the.gold

Catalan

oro del mundo. Spanish gold of.the world del món? Catalan of.the world

Turning now to Catalan and Spanish minimizers, these have been studied in depth by Vallduví (). As shown in () and () while Catalan minimizers are necessarily headed by ni ‘not even’, the particle is optional in Spanish minimizers.¹¹

¹¹ As already mentioned in Vallduví (: , fn. ), in some Catalan dialects the particle ni can be dropped (supposedly as a result of language contact with Spanish), with minimizers then displaying the distribution described for their Spanish counterparts.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   ()

a. No se sentia *(ni) una mosca. not . hear.. not a fly ‘One couldn’t hear the slightest sound.’



Catalan

b. No passava *(ni) un bri d’aire. not pass.. not a bit of.air ‘There wasn’t a breath of air coming through.’ c. No em mouré *(ni) un pam. not -self move.. not a palm ‘I won’t move an inch.’ (Adapted from Vallduví : , examples (), (), and ()) ()

a. No queda (ni) (una) gota de vino. not remains not a drop of wine ‘There isn’t a drop of wine left.’

Spanish

b. No tiene (ni) (una) pizca de gracia. not has not a pinch of grace ‘It isn’t a bit funny.’ c. No le toqué (ni) un pelo. not  touch. . not a hair ‘I didn’t touch him/her at all.’ (Adapted from Vallduví : , examples (), (), ()) Furthermore, ni-minimizers (and ni-maximizers) behave like NCIs. NCIs are typical of Negative Concord languages and, unlike PIs, they can occur in isolation as fragment answers to questions, (A0 ), and in preverbal position, (b, d). () a. Q: ¿Quién vino a la reunión? who came to the meeting ‘Who attended the meeting?’ A: *(Ni) un alma. not.even a soul ‘*(Not) a soul’ A0 :Nadie. n-body12 ‘Nobody.’ b. Q: Què has dit? what have. said ‘What did you say?’

Spanish

(minimizer)

(NCI)

Catalan

¹² Nadie and res have been glossed as ‘n-body’ and ‘n-thing’ in () and () because NCIs display an ambiguous behavior. When they are under the scope of negation (or other non-veridical operators in Catalan), they seem to be PIs, but when they occur preverbally or as fragment answers, they align with negative quantifiers. This behavior is typical of Romance NCIs (see Chapter  in this volume).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



  A: (*Ni) una paraula. not.even a word ‘Not a word.’ A0 : Res. n-thing ‘Nothing.’

()

oía. (minimizer) a. *(Ni) una mosca se not a fly . hear ‘Not the slightest sound could be heard.’ b. Nadie se oía. n-body . hear ‘Nobody could be heard.’ c. Ni un cèntim (no) ha costat. (minimizer) not a cent (not) has cost ‘It didn’t cost a red cent.’ d. Res (no) ha costat. n-thing (not) has cost ‘It cost nothing.’

(minimizer)

(NCI)

Spanish

(NCI)

Catalan

(NCI)

In Spanish, where NCIs cannot generally occur in non-negative contexts, ni-minimizers are as expected not felicitous in these contexts, but their ni-less counterparts are allowed, (). In Catalan, by contrast, NCIs can occur in non-negative contexts, but ni-minimizers cannot, (). Recall that the particle ni is optional with Spanish minimizers, but obligatory with Catalan minimizers. ()

Spanish a. ¿Acaso dijiste (*ni) palabra cuando debías? perchance say.. not word when must.. ‘Did you say a word when you should have?’ b. Si dice (*ni) una palabra, avísame. if says not a word warn..me ‘If he says a word, let me know.’ (Adapted from Vallduví : , example ())

()

ni un dit, per ell? Catalan *Que mouries Q move.. not a finger for him ‘Would you lift a finger for him?’ (Adapted from Vallduví : , example ())

The facts in () and (), which were already noticed by Vallduví (), can be straightforwardly accounted for by adopting an analysis of NCIs that assumes them to be PIs with a special syntactic requirement, namely that of carrying a syntactic polarity feature that has

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



to engage in an Agree chain with a licensing operator.¹³ Let us assume that NCIs are, unlike other kinds of PIs, specified with an uninterpretable polarity feature that needs to be checked by some particular polar operator. For Catalan NCIs, which can occur in negative contexts but also in non-negative ones, let us hypothesize that such a feature enters the derivation unvalued, (i.e. as [uPol: ]), and needs to be valued (and hence checked) by some polar operator (but not necessarily a negative one). For Spanish NCIs, by contrast, the feature would enter the derivation already valued (i.e. as [uPol:neg]), and would thus probe for a negative operator to do the checking. When occurring in isolation as fragment answers, or in preverbal position, the presence of an unchecked polarity feature would trigger the insertion of a Last Resort abstract operator (cf. Zeijlstra a; Espinal and Tubau b) to license the NCI. If this analysis is extended to Catalan and Spanish minimizers (and maximizers headed by ni), the particle ni can be claimed to encode a [uPol:neg] feature. Given that Catalan minimizers are obligatorily headed by ni, they are assumed to always be specified as [uPol: neg], thus being excluded from non-negative contexts such as (). The same happens with Spanish minimizers, which, as shown in (), cannot occur in non-negative contexts when headed by ni, but are fine in them when ni is absent. Furthermore, the presence of [uPol: neg] in ni allows minimizers to occur in fragment answers and in preverbal position, as it triggers a licensing Last Resort abstract negative operator.

Table . Catalan and Spanish maximizers and minimizers as different types of PI Type of PI NPIs

Maximizers Superstrong

ni per tot l’or del món (Cat.)/ni por todo el oro del mundo (Sp.)

Strong

en meses/años/siglos (Sp.)

Superweak

tot el temps del món (Cat.)/todo el tiempo del mundo (Sp.)

Superstrong

trigar la vida (Cat.)/tardar la vida (y media) (Sp.); cantidad (Sp.)

APIs

Superweak

una mosca; (una) gota de vino; (una) pizca de gracia; un pelo; un alma; palabra; mover un dedo (Sp.)

NPIs

Superstrong

ni-minimizers (Cat. and Sp.)

PPIs Type of PI

Minimizers

¹³ Espinal and Tubau (b) propose that NCIs are PIs that associate with a [uNeg] feature (Zeijlstra a) in a process of word syntax. Being PIs, NCIs can occur in non-veridical contexts, while the association with the [uNeg] feature makes them negation-dependent, and also able to occur in isolation as fragment answers, and in preverbal position. An unchecked [uNeg] feature (e.g. in NCIs in fragment answers or in preverbal position) triggers the insertion of a Last Resort abstract negative operator.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

Suleymanova and Hoeksema () report some Azerbaijani data that are reminiscent of what I have just discussed for the particle ni. In Azerbaijani, a minimizer headed by the particle heç requires to be licensed by clause-mate negation and patterns with nobody (while minimizers without heç are claimed to pattern with anybody/somebody).¹⁴ Further research should tell us whether minimizers (and maximizers) in other languages are also associated to specific particles, and whether these relate to negation in ways that are similar to the ones described for Spanish and Catalan. To sum up, as shown in Table ., Catalan and Spanish maximizers correspond to different types and subtypes of PIs. By contrast, minimizers without ni are superweak APIs, while minimizers and maximizers with ni are superstrong NPIs and, like NCIs, can occur as fragment answers and in preverbal position in the absence of an overt licensor.

.. A       E, C,  S

.................................................................................................................................. Vulgar minimizers (referred to as a class by Postal (b) under the label ‘SQUAT’), (), have been discussed in McCloskey (), Horn (c), Postal (b), and, more recently, De Clercq () for English, but similar examples exist in Catalan and Spanish, () and (). ()

SQUAT ¼ squat, fuck-all, beans, crap, dick, diddley, diddley-poo, diddley-squat, jack, jack-shit, jack-squat, piss-all, poo, shit, shit-all, sod-all, bugger-all, naff-all, crap-all (De Clercq : , example ())

()

Catalan: una merda ‘a shit’, un carall ‘a penis’, un colló ‘a testicle’, un cagarro ‘a turd’

()

Spanish: una mierda ‘a shit’, un carajo ‘a penis’, tres cojones ‘three testicles’, un mojón ‘a turd’

Vulgar minimizers in English can occur both in negative and affirmative sentences, (a), as well as in questions, (b), and in fragment answers. This is also the case in Catalan and Spanish, ().¹⁵

¹⁴ Suleymanova and Hoeksema () analyze heç as an inherently negative particle that concords with the verb, which carries a negative suffix. ¹⁵ With some predicates (e.g. entendre (Cat.) / entender (Sp.)), vulgar minimizers require licensing by negation in declaratives, (i), but then can occur in non-negative contexts, (ii), and in fragment answers, (c). See also fn. . (i) *(No) entiendo {un carajo/ una mierda} del tema. penis a shit of.the topic not see. a ‘I don’t understand squat about what you say.’

Spanish

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   ()

a. I (don’t) know squat about physics. b. Does he know squat about physics? c. Speaker A: What does she know about physics? Speaker B: Squat.

()

a. (No) m’importa una merda. not to.me.matters a shit ‘I don’t give a shit.’



English

(Catalan)16

b. T’ importa una merda aquest tema? to.you matters a shit this topic ‘Do you give a shit about this topic?’ c. Speaker A: Què has entès de la xerrada? what have. understood of the talk ‘How much of the talk have you understood?’ Speaker B: Una merda. a shit ‘Squat’ As shown in (), English vulgar minimizers do not express sentential negation (as diagnosed by Klima’s () classical tests) in the absence of the negative marker. This is also the case for Spanish (and Catalan), (). ()

a. She knows squat about the topic, *does she? / doesn’t she? b. She knows squat about the topic, and *neither does John / so does John.

()

Me importa una mierda, y a Juan le importa una mierda to.me matters a shit and to Juan to.him matters a shit *tampoco/ también. either too ‘I care squat, and John cares squat *either / too.’

Postal (b), inspired by Déprez’s () work, claims that the structure of vulgar minimizers involves an incorporated cardinal numeral zero, as in (). ()

[DP [D zero] þ [N squat]]

This would explain why vulgar minimizers can occur both in negative and non-negative contexts, but are not diagnosed as conveying sentential negation in () and (). Yet, the

(ii)

¿Has entendido {un carajo/ una mierda} del tema? have.. understood a penis a shit of.the topic ‘Have you understood squat about the topic?’ ¹⁶ Similar data are attested in Spanish.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

optionality of negation in (a) and (a) leads us to entertain the possibility that lexical ambiguity exists among vulgar minimizers (cf. Herburger  for lexical ambiguity in NCIs), so that they can be superweak APIs with the licensing conditions established in Table ., but also lexical items with the structure in ().¹⁷ A more accurate substantiation of this claim, nonetheless, is left as further research.

.. S  

.................................................................................................................................. In this chapter I have explored the nature of maximizers and minimizers in English, Catalan, and Spanish and have shown that in spite of not forming a homogeneous class, they display properties of different types and subtypes of PI. I have argued that they can be classified according to their potential (anti-)licensors using a concentric model of polarity that combines DE and non-veridicality as the relevant (anti-)licensing conditions. I have also investigated the connections of minimizers with other items in the polarity landscape such as negative quantifiers and NCIs. For English, it has been observed that API- and NPI-minimizers behave as canonical PIs under the scope of negation and other non-veridical operators, but as negative quantifiers when the minimizer is adjacent to negation. As it has been argued in the literature that minimizers contain a Focus head in their structure that accounts for their emphatic nature, it has been suggested that they can undergo QR when adjacent to negation to take sentential scope. Further research should determine whether the parallelism between English minimizers and negative quantifiers is also observed in other languages. Spanish and Catalan minimizers, by contrast, have been shown to align with NCIs when headed by the particle ni, which is optional in Spanish, but obligatory in Catalan. Like NCIs, ni-minimizers and maximizers can occur in isolation and in preverbal position, and cannot be used in questions and conditionals with a non-negative meaning. I have argued that ni carries a syntactic feature [uPol:neg] that makes minimizers and maximizers negation-dependent (i.e. superstrong NPIs), and is responsible for triggering a Last Resort abstract negative operator that allows them to occur in fragment answers and preverbally. The distribution of vulgar minimizers has also been briefly addressed. In the three languages under study, vulgar minimizers can optionally be licensed by negation, as well as by other non-veridical operators, and can occur in fragment answers. I have attributed the optionality of negation to the idea that vulgar minimizers are lexically ambiguous between DPs with an incorporated zero numeral (Postal b), and canonical superweak APIs. While the zero-incorporated structure approach to vulgar minimizers is compatible with these expressions not needing licensing by negation, and with their contributing of negative meaning despite negation being diagnosed as non-sentential, the fact that they can also be licensed by negation seems to require their treatment as PIs, as well.

¹⁷ If the lexical ambiguity analysis of vulgar minimizers is on the right track, the Spanish data in fn.  can be related to certain predicates always selecting the API lexical entry for vulgar minimizers rather than the zero-incorporated lexical entry. Why certain predicates should have certain preferences for one entry or another is an open research question.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



All in all, this chapter has tried to show that minimizers and maximizers fit well into the class of PIs if this is assumed to be broad enough so as to allow for a number of distinct types and subtypes. In this chapter, types of PI have been established on the basis of what elements can act as potential (anti-)licensors. In addition, I have also argued that it is possible to establish interesting connections between minimizers (and ni-headed maximizers) to other kinds of polarity-related lexical items. This should lead us to a better understanding of the different items that form the complex constellation of polarity sensitive elements.

A

.................................................................................................................................. This research has been funded by a research grant awarded by the Spanish Ministerio de Economía y Competitividad (FFI--P), and by a grant awarded by the Generalitat de Catalunya to the Centre de Lingüística Teòrica (SGR). I would like to thank two anonymous reviewers for comments and suggestions on how to improve the manuscript. All errors remain my own.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ......................................................................................................................

  ......................................................................................................................

 

. I

.................................................................................................................................. A universal property of natural language is that every language is able to express negation. Every language has some device at its disposal to reverse the truth value of the propositional content of a sentence. However, languages may differ significantly as to how they express this negation. Languages may, for instance, use different categorial elements in order to express negation; English, for example, apart from having two negative markers not and n’t as shown in (), can also use negative indefinites (nobody, nothing, never, . . . ), as is illustrated in (). Cross-linguistically not every language displays such negative indefinites; Hindi, for instance, lacks them. ()

a. John did not leave b. John didn’t leave

()

a. Nobody left b. Mary never went there c. I’d never do that

The central topic of this article are these negative indefinites. Under standard Montegovean semantics, negative indefinites like no are taken to be negative quantifiers with the denotation in (a); negative indefinites like nobody have a semantics as in (b): ()

a. λXλY. ¬∃x.[X(x) & Y(x)] b. λX. ¬∃x.[Person(x) & X(x)]

The semantics in () gives rise to at least two different questions or problems. First, if a negative indefinite consists of a negation and an existential quantifier, the question arises whether negative elements consisting in the same way of a negation and a universal quantifier would be possible as well. In other words, can there be negative elements that have the meaning in ()?

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ()



λXλY. ¬∀x.[X(x) → > Y(x)]

Such elements would have a meaning like ‘not all’, but in natural language such elements are not attested. Such meanings cannot be expressed by single lexical items. But why would this be the case? If a negative indefinite may consist of a negation and an existential quantifier, why can’t there be single lexical items consisting of a negation and a universal quantifier? Second, it turns out that negative indefinites in Double Negation languages, that is languages that do not display Negative Concord, which should therefore be considered to carry a semantic negation of their own may give rise to so-called split-scope readings where the negative part of a negative indefinite outscopes a scope-taking element that in turn outscopes the indefinite part. The German sentence in () below has as its most salient reading that it is not the case that you are required to wear a tie. ()

Du musst keine Krawatte anziehen You must no tie wear a. ‘It is not required that you wear a tie’ b. ‘There is no tie that you are required to wear’ c. ?‘It is required that you don’t wear a tie’

German ¬ > must > ∃ ¬ > ∃ > must must > ¬ > ∃

This, again, suggests, that negative indefinites in such languages cannot straightforwardly be taken to be negative quantifiers with a meaning as in (). In this overview chapter, I discuss these two challenges. First, in section ., I address the so-called NALL problem, that is the question why negative universals are never attested in natural language. Then, in section ., I discuss the phenomenon of split-scope and the way it has been analyzed over time. Section . concludes.

. T NALL 

..................................................................................................................................

... The problem The overall meaning contribution of a negative indefinite is that of a negated existential (or a universal quantifier outscoping negation, which is truth-conditionally equivalent). Negative quantifiers, just like universal quantifiers in positive environments make strong meaning contributions. This contrasts with existential quantifiers / indefinites in positive contexts, which make a weak meaning contribution. The examples in () are informationally strong, and easily falsifiable, whereas the example in () is weak and easily verifiable. ()

a. I saw every girl in the city b. I saw no girl in the city

()

I saw some/a girl in the city

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

An equally weak example as in () would be a negated universal, as in (). ()

I didn’t see every girl / all girls in the city

However, while both positive and negative indefnites have been widely attested, as have positive universals, negative universals have not been attested: There are no languages with single lexical items meaning ‘not all’ or ‘not every’. The observation that no language in the world has a single lexical item with the meaning ‘not all’ or ‘not every’ is well established and goes back to the works of Thomas Aquinas. Horn (, , ) extends the observation to a ban on lexicalization of the so-called O-corner in the Square of Opposition (see also Levinson ). The Square of Opposition is a diagram depicting the four major types of propositions under Aristotelian logic: universal affirmatives (A), universal negatives (E), existential affirmatives (I), and existential negatives (O), each of them exemplified below (the abbreviations stem from Latin AffIrmo (‘I assert’) and nEgO (‘I deny’)): ()

a. b. c. d.

Universal affirmative (A): Universal negative (E): Existential affirmative (I): Existential negative (O):

every car is red no car is red some car is red Some car is not red = Not every car is red

Boethius (–) formalized this in his Square of Opposition. He included each of the four basic sentence types in the corners of the square, A and I on the affirmative side of the square, and E and O on the negative side, and described the logical relations between them. ()

Square of Opposition A

E

Contradictions

I

< -Subcontraries ->

O

The observation of Horn and others boils down to an alleged ban on lexicalizing the O-corner of the Square of Opposition. English quantifiers like every, some, and no can be seen as lexicalizations of the A, I, and E corner, respectively. But English lacks a single quantifier meaning ‘not all’ (the unattested word nall, hence the name of the problem).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 



Similarly, whereas English has both, either, and neither, there is no noth, a word that would mean ‘not both’. Propositional connectives can also appear in the Square of Opposition as they are subject to the same entailment relations. Again, and (A), (inclusive) or (I), and nor (E) exist, but nand (with the intended meaning ‘not and’) doesn’t. This is not a particular quirk of English. Languages in general have no problem in lexicalizing the A, I, and E corners, albeit that the E-corner is not always lexicalized; not every language exhibits negative indefinites or negative coordinators. However, across languages the O-corner is systematically excluded from lexicalization (see Jaspers  and references therein for more overview and discussion).

... Synchronic solutions The question why it is impossible to lexicalize the O-corner of the Square of Opposition has received several answers, but has never been fully satisfactorily solved. With the exception of Hoeksema (b), most existing solutions to the nall-problem allude to particular synchronic, grammatical, or logical mechanisms to rule out O-lexicalization. One grammatical principle that can be alluded to in order to exclude O-lexicalization is blocking: the existence of the lexicalization of the I-corner blocks lexicalization of the O-corner, as argued for by Horn (, , ). His argument is that whereas I- and O-type sentences are semantically different, their pragmatic contributions are similar. To see this, let’s take into consideration the joint meaning contribution of the assertions and the scalar implicatures of the following two examples: () Assertion Some car is red Implicature Not every car is red Joint meaning contribution: Some but not every car is red () Assertion Not every car is red Implicature Some car is red Joint meaning contribution: Some but not every car is red Since the joint meaning contributions made by the corresponding I- and O-type sentences are the same, natural language needs only to exploit one. Since negative expressions are always marked in comparison to their positive counterparts, for Horn the possible existence of I-type terms blocks the existence of O-type terms. If a language is to lexicalize one of the I/O-corners it should be the I-corner. As Hoeksema (b) points out, pragmatic equivalence is much weaker than semantic equivalence. One can easily utter the assertion in () about a particular subset of cars without knowing anything about the colours of the other cars. But then the pragmatic equivalence of () and () is disrupted. Moreover, if conveying I-type sentences blocks conveying O-type sentences, the question emerges why utterances containing expressions like ‘not . . . every’ or ‘not . . . all’ are still fine. In fact, their appearance is abundant. For these and other reasons, other scholars have looked for other kinds of arguments to account for the ban on O-lexicalization. Most notably Jespersen () and Seuren (), following up on work by De Morgan (), argued that the Square of Opposition should

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

actually be replaced by a triangle where either the O-corner would be dropped altogether or where it would conflate with the I-corner. Jespersen proposes a tripartition of logical operators (including only every, some, and no), which therefore only require three corners. Naturally, if there is no O-corner left, there is no O-problem to begin with. Alternatively, other ways of designing the square (cf. Löbner , ) or other geometric representations (such as hexagons) have been proposed as well (see Van der Auwera ; Jaspers ; Horn ; see also Béziau and Gianfranco  for overview and discussion on geometrical alternatives to the Square of Opposition). Under such richer geometrical representations there are still O-corners, but these are no longer the only ones that are excluded from lexicalization. At the same time, as Jaspers () mentions, even if there is no O-corner anymore left, the question still remains open why logically accessible meanings, such as ‘not all’, ‘not both’, or ‘not and’ still cannot be lexicalized. Again, this point becomes especially relevant in the light of Hoeksema’s observation that the complex construction ‘not . . . every’ is highly pervasive.

.. Diachronic solutions Attempts that have aimed to provide a synchronic explanation for the NALL-problem have thus far not been able to fully solve it. For this reason, one may instead opt for a diachronic approach. Maybe there is no grammatical or logical reason why elements living in the O-corner cannot be lexicalized, but the conditions under which this lexicalization would emerge are never met. As a starting point, consider that every negative expression is marked (as clearly demonstrated by Horn ) in comparison to their positive counterparts. Negative quantifiers, therefore, should be the result of some process that made their positive counterparts negative. Hoeksema (b) discusses two such processes: lexical merger of a negative marker and an indefinite or a Negative Polarity Item (NPI), or reanalysis of indefinites/NPIs. Examples of the former are the negatively marked series of negative indefinites in languages like English. Examples of the latter can be attested in French, where Negative Concord Items (NCIs), such as personne ‘nobody’ or rien ‘nothing’, lack any reflection of negative morphology. Now let us look at the scenarios under which these processes can take place. First note, that under the reanalysis process, an element that is initially intended to strengthen the negation (cf. Kiparsky and Condoravdi ) is first reanalyzed in such a way that it becomes a semantically bleached NPI. This NPI, always co-occurring with negation is then felt to be a negative element (i.e. a negative marker or a Negative Concord Item / neg-word) leaving the original negative marker a superfluous element. Examples for French are pas and personne that originally meant footstep or person. Later on these got grammaticalized as minimizing NPI (meaning something like a single bit or anybody) and finally they became a negative marker and an NCI respectively. For a universal quantifier to undergo the same process, it should first appear in a stage where (i) it is used to strengthen a negation and (ii) only or mostly appears in negative sentences. However, as pointed out by Hoeksema (b), in order to strengthen a negation, a universal quantifier should outscope negation rather than scope under it.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 



Universals scoping under negation rather yield weakened readings. But if this universal may not appear under the scope of negation, it cannot be lexicalized as a universal quantifier scoping under negation. Hence, the two conditions (i)–(ii) that would require universals to enter the first stage of reanalysis into negative universals are in conflict, and reanalysis of this type is thus predicted not to take place. The only candidate left for lexicalizing negative universals (and other inhabitants of the O-corner) over time would then be lexical merger. Hoeksema excludes this option by arguing that it would be very hard for the negative marker expressing sentential negation to immediately precede the universal quantifier, a necessary condition for a negative element to undergo lexical merger with a universal. Neither in SVO nor in SOV languages would such a negative marker easily appear in a string-adjacent position next to the universal. However, it is very well possible that a negative prefix that can be attached to verbs may be attached to quantifying elements. That something of the kind must have happened can be straightforwardly assumed, as string-adjacency between a negative marker and an indefinite is also a necessary condition for lexical merger of negative indefinites, a process that frequently occurred over time. Consequently, it is not clear yet whether a diachronic approach may fare better in explaining the universal absence of lexicalized negated universal quantifiers, though see Zeijlstra (b), who argues that most universal quantifiers under negation are focused, whereas only unfocused universal quantifiers preceded by negation can be candidates for lexical merger.

.. S 

..................................................................................................................................

... The phenomenon As indicated in section ., one way to express sentential negation other than by means of negative markers, exploited in various languages, is by using negative indefinites. We already showed this for English in (). Also, in Dutch and German, sentential negation can be expressed by means of a sole negative indefinite, in either preverbal or postverbal position. ()

a. Niemand ziet haar Nobody sees her ‘Nobody sees her’ b. Zij ziet niemand She sees nobody ‘She sees nobody’

()

a. Keiner seht sie Nobody sees her ‘Nobody sees her’ b. Sie seht keinen She sees nobody ‘She sees nobody’

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

These facts strongly suggest that negative indefinites in these languages are semantically active negative quantifiers. In this section, I introduce, basing myself again on Dutch, German, and to a lesser extent English, a phenomenon that casts doubt on the view that negative indefinites in Double Negation languages are negative quantifiers. Take examples () and (), from German and Dutch, respectively (discussed in Rullmann a; Penka , among others): () Du musst keine Krawatte anziehen tie wear You must no a. ‘It is not required that you wear a tie’ b. ‘There is no tie that you are required to wear’ c. ?‘It is required that you don’t wear a tie’ () Ze mogen geen eenhoorn zoeken no unicorn search They may a. ‘They are not allowed to search any unicorn’ b. ‘There is no unicorn that they are allowed to search’ c. ?‘They are allowed not to search a unicorn’’

German ¬ > must > ∃ ¬ > ∃ > must must > ¬ > ∃ Dutch ¬ > may > ∃ ¬ > ∃ > may may > ¬ > ∃

In both examples, three readings are available: one where the entire negative indefinite takes wide scope with respect to the modal verb (¬ > ∃ > must/may), one (slightly less available) reading where the entire negative indefinite takes narrow scope (must/may > ¬ > ∃), and a so-called split-scope reading where the negative part of the negative indefinite outscopes the modal but where the indefinite part still scopes under the modal (¬ > must/ may > ∃). Whereas the first and second readings could simply reduce to the wide or narrow scope of the negative quantifier, the third reading cannot do so. For instance, in () the wide scope reading of the negative indefinite says that there is no particular unicorn that they may search, and the narrow scope readings say that they are allowed not to find any unicorn. The split-scope reading is compatible with neither of these readings; under this reading they are simply not allowed to search any unicorn. Independent evidence for the existence of these readings, showing that they are not simply entailed readings, can be found in example (), discussed in Penka (). German brauchen, being an NPI, must scope under negation. At the same time, when appearing in an existential construction with expletive es ‘there’, an indefinite embedded under a modal can only take narrow scope with respect to it. Consequently, both the wide scope and narrow scope readings of the negative indefinite in example (b) and (c) respectively are ruled out. Still, the sentence is grammatical with a split-scope reading. This finding provides independent evidence for the existence of separate split-scope readings.

() Es braucht kein Arzt anwesend zu sein. no physician present to be There needs a. ‘It is not required that there be a physician present’ b. *‘There is no physician who is required to be present’ c. *‘It is required that there be no physician present’

German ¬ > need > ∃ ¬ > ∃ > need need > ¬ > ∃

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 



Split-scope effects do not appear only in combinations with modals and negative indefinites. Object-intensional verbs also invoke split-scope readings (although the narrow-scope reading is independently ruled out; see Zimmermann ), as is shown in examples () and (), again for German and Dutch. () Perikles schuldet Socrates kein Pferd Socrates no horse Perikles owes a. ‘Perikles is not obliged to give Socrates a horse’ b. ‘There is no horse that P. is obliged to give to Socrates’ c. *‘Perikles is obliged not to give Socrates a horse’ () Hans zoekt geen eenhoorn Hans seeks no unicorn a. ‘Hans is not trying to find a unicorn’ b. ‘There is no unicorn Hans is trying to find’ c. *‘Hans is trying not to find a unicorn’

German ¬ > owe > ∃ ¬ > ∃ > owe owe > ¬ > ∃ Dutch ¬ > seek > ∃ ¬ > ∃ > seek seek > ¬ > ∃

Also, non-negative idiomatic expressions containing an indefinite allow split-scope readings when these indefinites are negated, which is interesting as idiomatic expressions normally form a fixed unit and cannot be intervened (cf. Koeneman and Zeijlstra ). Generally, German and Dutch idioms involving an indefinite are negated by replacing the indefinite with a negative indefinite. The negation then applies to the entire idiom that contains the indefinite (see ()–()). () a. Hans hat mir einen Hans has me a ‘Hans has fooled me’

Bären aufgebunden bear up-tied

German

b. Hans hat mir keinen Bären aufgebunden Hans has me no bear up-tied ‘Hans hasn’t fooled me’ ()

a. Hij heeft een scheve a diagonal He has ‘He made a mistake’

schaats skate

gereden ridden

Dutch

b. Hij heeft geen scheve schaats gereden He has no diagonal skate ridden ‘He didn’t make any mistake’ Hence, a paradox arises. On the one hand, the conclusion that negative indefinites in Double Negation languages are negative quantifiers seems sound. But such negative indefinites do not manifest the precise behavior one might expect if they are indeed negative quantifiers. Either one is thus forced to assume that these negative indefinites are indeed negative quantifiers but that their deviant behavior (i.e. their ability to give rise to split-scope readings) follows from something else, or one has to give up the idea that they

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

are plain negative quantifiers. Under the latter perspective, negative indefinites in the Germanic languages discussed above are in one way or another decomposed into negations and existentials that get jointly spelled out. Below I will discuss the two approaches in a bit more detail. After that I will introduce work that argues that scope-splitting is not unique to negation and should be thought of in broader terms.

... The negative quantifier approach One of the first representatives of the negative quantifier approach is Geurts (a), who argues that negative indefinites, at least in Double Negation languages, are negative quantifiers, which may not only quantify over individuals but can also quantify over kinds (in the sense of Carlson ). To see this, take () (after Geurts a): ()

Ich suche keinen Putzmann I seek no cleaning-man ‘I do not look for a cleaning man’

Whereas for Geurts both the narrow- and wide-scope readings of () involve quantification over individuals, he also takes the sentence to have a reading where cleaning man refers to a kind. In this case, the sentence would mean that there is no kind-element cleaning man such that I seek this kind, which means that ‘I am not a cleaning-man-seeker.’ That reading is equivalent to the split-scope reading: I am not a cleaning-man-seeker, that means that it is not the case that I am seeking a cleaning man. If this line of reasoning is correct, no additional assumptions have to be made in order to account for the existence of split-scope readings, and the null hypothesis that all negative indefinites in Double Negation languages are negative quantifiers can be maintained. However, several scholars, most notably Penka (), have pointed out several problems for this approach, both conceptually and empirically (see de Swart b for a more thorough discussion of these problems). As pointed out by Penka (), Geurts cannot always simply appeal to the notion of abstract individual or natural kinds as used in Carlson (). To account for split readings, in some cases very specific and strange kinds would have to be assumed. For instance, to get the paraphrased reading of (), Geurts would have to appeal to the kind ‘student who attended Arnim’s lecture yesterday’. () Ich suche keinen Student, der gestern in Arnims Vorlesung war German seek no student who yesterday in Arnim’s lecture was I ‘I’m not looking for a student who attended Arnim’s lecture yesterday.’ Another problem for this analysis is the fact that kein can combine with numerals while scope splitting is still possible. It remains unclear how Geurts’s account could deal with a sentence such as () under the reading paraphrased, as two cars does not refer to a particular kind in terms of Carlson ().

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  () Wir müssen keine zwei Autos haben We must no two cars have ‘We don’t need to have two cars.’



German

An alternative account within the negative quantifier approach is de Swart (b). With Geurts she assumes that some special kind of quantification is responsible for scope splitting of negative indefinites. But rather than assuming quantification over abstract individuals, which causes several of the above-mentioned problems, de Swart (b) employs higher-order quantification. She argues that scope splitting occurs when kein quantifies over properties and proposes that there is an additional lexical entry for kein according to which kein is a negative quantifier over properties. The denotation of kein Buch (no book) in () would then be (), yielding the split-scope reading in (): () Hanna sucht kein Buch Hanna seeks no book

German

() [[ kein Buch ]] = λw.λP.¬∃Q (Q=λw’λy.(Book’w’(y)) & P(Q)) () no book (seek) (hanna) ¬∃P(P= λw’.λy.(Book’w’ (y)) & Seek’(h,P)) = ¬Seek’(h, λw’.λy.(Book’w’(y)) ‘Hanna is not a book seeker’ = ‘Hanna doesn’t seek a book’ But, as Penka () points out, there are reasons to believe that higher-order quantification is not what is responsible for scope splitting. First, such an analysis cannot derive intermediate scope readings of the indefinite for sentences with two scope-bearing elements besides negation and the indefinite, that is readings where the negation takes widest scope and the indefinite takes scope in between the two operators. This is the case because the higherorder interpretation of kein invariably gives the indefinite narrowest scope. Empirically, this appears to be too strong. For example, () does indeed have the reading paraphrased, in which negation takes widest scope and the indefinite scopes in between kann ‘can’ and wollen ‘want’. This sentence comes along with the reading that it is impossible that there is a Norwegian such that Julia wants to marry him. () Julia kann keinen Norweger heiraten wollen Julia can no Norwegian marry want ‘It’s not possible that there is a Norwegian that J. wants to marry’

German ¬>can>∃>want

... The decomposition approach An alternative way of accounting for split-scope effects is decomposition. Under this approach the two ingredients of a negative indefinite, the negation and the indefinite, are

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

different (lexical) elements that in certain configurations may jointly be spelled out as a single negative indefinite. The approach was initiated by Jacobs () and Rullmann (a) for whom negative indefinites, such as German kein or Dutch geen, underlyingly consist of a plain negative marker and a plain indefinite. Only under PF adjacency can the two be jointly spelled out as a single morphological word, following a rule like the following (for German and for Dutch). ()

nicht þ ein → kein nicht þ een → geen

Again, the split-scope reading can emerge if nicht/niet and ein/een are adjacent at PF while structurally separated by some additional scope-taking element at LF. The three readings that () may yield come about as in ( a–c) (where the modal is interpreted in its base position). ()

Ze mogen geen verpleegkundige ontslaan They may no nurse fire a. [Zij mogeni [niet een verpleegster [ontslaan ti]]] b. [Zij mogeni [niet een verpleegster ontslaan] ti]] c. [Zij mogeni niet [[een verpleegster ontslaan] ti]]

¬ > ∃ > may may > ¬ > ∃ ¬ > may > ∃

This decomposition approach, in its various forms, circumvents the semantic problems of split-scope effects by taking the spelling out of negative indefinites to be a PF phenomenon, but it comes with challenges of its own. For one, such analyses can apply only to OV languages, languages where a VP-external negative marker may appear immediately left of a VP-internal object. In VO languages, such negative markers would precede the verb, but the object would follow them. Then the negation and the object are no longer string-adjacent. However, some speakers of English, a clear VO language, do allow split-scope readings as well (cf. Iatridou and Sichel ). In () there is no way in which this split-scope reading could have been derived from PF adjacency of the negative marker and the indefinite: ()

You have to do no homework today

To solve this problem, Zeijlstra () proposes a mixture of the negative quantifier approach and the decomposition approach. In short, he posits that the negation and the indefinite merge into a negative quantifier that can further undergo quantifier raising. This process leads to two copies of the [ [NEG] þ [INDEF] ] treelet. Consequently, one of these copies may be spelled out at PF as a negative indefinite along the lines of Jacobs (), Rullmann (a), and also Penka (), whereas at the level of LF, nothing forbids the negation from being interpreted in the higher copy and the indefinite in the lower copy. The underlying LF of example () is then as in (), where the lower copy gets jointly spelled out as no homework at PF: ()

You [[[] []] homework] have to do [[[] []] homework]

On the one hand, under Zeijlstra’s account, the decomposition approach can be extended to all languages, rather than apply only to OV languages. On the other hand, this account

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 



crucially relies on the availability of negative quantifiers to undergo quantifier raising. This assumption is not uncontroversial. Von Fintel and Iatridou () have shown that negative indefinites cannot scope over other quantifiers, as the following examples demonstrate for English: () Everybody touched no dessert

∀ > ¬∃; *¬∃ > ∀

Thus, Zeijlstra’s account can only be successful if some other mechanism is responsible for ruling out the possibility of the negative indefinite to outscope the universal quantifier in sentences such as example (). Another account that falls under the decomposition approach, proposed by Penka (), tries to connect the behavior of neg-words in Negative Concord languages with the behavior of negative indefinites in Double Negation languages. For her, all negative indefinites, in both Negative Concord and Double Negation languages, are semantically non-negative. Penka draws a parallel between Negative Concord and split-scope readings, and argues that the same process underlies both phenomena. Penka adopts Zeijlstra’s approach to Negative Concord where NCIs are taken to be semantically non-negative indefinites, which carry an uninterpretable negative feature [uNEG] that needs to be checked against a potentially covert negative operator carrying [iNEG]. Penka argues that in Double Negation languages like Dutch and German, the same process is going on, the only difference being that multiple Agree between one negation operator and multiple NCIs is not allowed. In these languages, a negative operator can check off only one negatively marked indefinite. Thus, every negative indefinite is semantically nonnegative and carries a feature [uNEG], but it also needs to have its feature checked against a unique abstract negative operator Op¬[iNEG]. If two negative indefinites show up in the sentence, each negative indefinite must be licensed by a separate Op¬[iNEG]. Penka then derives split-scope readings by having the abstract negative operator outscope the intervening operator, which in turn outscopes the indefinite DP, as illustrated for example in (). Penka argues that the position of the abstract negative operator Op¬ should be PF-adjacent to the position of the NCI. In OV languages, like Dutch and German, that means that the Op could occupy a VP-external position, with the NCI appearing VP-internally, as was the case for Jacobs and Rullmann. ()

. . . dass Du keine Krawatte anziehen musst . . . that you no tie wear must [dass Du [OpNEG[iNEG] [[keine[uNEG] Krawatte anziehen] musst]]] ‘It is not required that you wear a tie’

German

¬ > must > ∃

The advantage of Penka’s account is that it adopts an independently motivated mechanism to account for split-scope readings, and Penka does not have to make any particular claims about the behavior of negative indefinites in Double Negation languages. However, this analysis also faces some problems. First, Penka takes every language to exhibit formal negative features, along the lines of Zeijlstra. However, Zeijlstra’s analysis is not directly compatible with this extension. Under his approach, the difference between Negative Concord and Double Negation languages is that negative indefinites in Double Negation languages have no formal negative features at all at their disposal (Zeijlstra a, a,

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

b, ); for Zeijlstra, formal negative features can be acquired only in Negative Concord languages. Moreover, it is not clear how the cross-linguistic variation with respect to the possibility of (not) being subject to multiple Agree can be lexically encoded. Another problem for Penka may be that her analysis does not readily extend to VO languages, even though VO languages, like English, may exhibit split-scope effects.

... Split-scope and negative indefinites All analyses of split-scope discussed thus far have been designed to account for split-scope readings yielded by negative indefinites. However, it has also been observed that splitscope readings may also be triggered by other quantifiers that are not upward monotonic. Take the following examples by Hackl () and Rullmann (a), quoted in Abels and Marti (): () At MIT one must publish fewer than three books in order to get tenure ‘At MIT one must publish at least n books in order to get tenure, and n is less than three’ (Hackl ) () How many books does Chris want to buy? ‘What is the number n such that Chris wants it to be the case that there are n books that he buys?’ (Rullmann a) That split-scope readings may emerge in comparatives and comparative quantifiers is well known since Hackl (), de Swart (b), and Heim () (see also Alrenga and Kennedy ). The question that arises is whether these two types of split-scope phenomena, that is split-scope phenomena with respect to negation and with respect to other non-monotonic operators, such as comparative quantifiers, reflect different grammatical phenomena or can be said to follow from a single, unified analysis of split-scope phenomena. One reason, put forward by Jacobs () and Penka () for a non-unified approach to split-scope is that negative indefinites can also give rise to split-scope readings across universal quantifiers with a particular hat contour (cf. Büring ). For many speakers of German () receives a reading where not every doctor has a car. ()

/JEDER Arzt hat KEIN\ Auto every doctor has no car ‘Not every doctor has a car’

Crucially, however, such split-scope readings do not emerge with other non-upward monotonic quantifiers: ()

??

/JEDER Arzt hat weniger als DREI\ Autos every doctor has less than three cars ‘Every doctor has at least n cars, and n is less than three’

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ()



*/JEDER Arzt hat GENAU\ drei Autos every doctor has less three cars ‘Every doctor has at least three cars’

Since examples like () are not available for every speaker of German (unlike the earlier discussed examples of split-scope involving negative indefinites and modals and intentional operators), Abels and Marti () take split-scope involving the hat contour to be an independent phenomenon from other types of split-scope, thus paving the way for a unified approach for split-scope involving negative and other non-monotone increasing quantifiers, such as comparative quantifiers and numerals. For them, a negative quantifier like German kein is a negative quantifier over choice functions, which scopes from its raised position. The noun phrase is interpreted below where the trace of the quantifier leaves a choice function variable that takes the denotation of the noun as its argument. Since Abels and Marti take common noun denotations to be indexed by a world, where this index can be bound by a higher intensional operator, the impression of a split-scope reading is the result. The split-scope reading of (), repeated as (a), results from the LF in (b) (where @ stands for the actual world). ()

a. Du musst keine Krawatte anziehen You must no tie wear ‘It is not required that you wear a tie’

German ¬ > must > ∃

b. ¬∃f CF(f) & ∀w’R@, you wear f(tiew’) in w’ What remains unclear, however, is why speakers who accept split-scope readings with the hat contour systematically accept split-scope readings across intensional operators, something that is not immediately clear if the two phenomena are truly independent.

... Summing up So far we have seen that negative indefinites, at least in certain languages, give rise to splitscope readings, that is readings where the negation outscopes a particular scope-taking element that in turn outscopes the indefinites. Whether split-scope readings invoked by negative indefinites constitute a grammatical phenomenon of their own, or whether they are part of a bigger cluster of split-scope phenomena all involving non-upward monotonic quantifiers (see also Alrenga and Kennedy  for discussion) is a question that is currently subject to research.

. C

.................................................................................................................................. Negative indefinites, or elements that are traditionally referred to as such, show different behavior with respect to what would be expected if they were indeed plain negative quantifiers. Such differences involve their ability to yield split-scope readings, or the lexical absence of their mirror image: negated universal quantifiers.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

In this chapter, I have tried to sketch the analyses of these phenomena that are currently on the market. As of yet, there is no clear consensus in the field what the right direction would be to account for split-scope readings or how to solve the nall-problem. Each analysis comes with certain advantages and certain drawbacks, and each analysis is based on assumptions that may not be trivial or theory-neutral. What they all share, though, is the view that what sometimes simply looks like a plain negative quantifier, at closer inspection behaves in many unexpected and intricate ways, making the study of negation and negative quantifiers as fascinating as it is.

A

.................................................................................................................................. I am grateful to two anonymous reviewers, the editors of this volume, and everybody with whom over the years I have discussed aspects of the issues addressed in this chapter.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ......................................................................................................................

   ......................................................................................................................

 

.. I

.................................................................................................................................. T chapter discusses the interaction of negation with fragment answers, or subsententials—structures smaller than a clause which provide answers to (possibly implicit) questions, as in (). ()

a. What did John eat?—[DP Chips]. b. Where did Mary go?—[PP To Paris]. c. John was speaking to someone.—Yeah, (to) Mary.

The chapter will focus on two theoretically interesting ways in which negation can interact with fragment structures. The first phenomenon of interest is the use of negative concord items (NCIs, ()) and negative polarity items (NPIs, ()) as fragment answers. () a. ¿Qué comiste? —Nada. what eat-. n-thing ‘What did you eat?—Nothing.’

Spanish

b. Nani-o mita no? —Nani-mo. what- saw  what- ‘What did you see?—Nothing.’ c. Cine nu a venit? —Nimeni. who not has come n-body ‘Who didn’t come?’ i. ‘Nobody.’ (= nobody came) ii. ‘Everybody.’ (=nobody didn’t come) ()

Japanese (Watanabe : ) Romanian

(Fălăus and Nicolae : )

a. What did he bring to the picnic?—*Any wine. b. What didn’t he bring to the picnic?—?/%Any wine.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

Fragments such as (a,b) are grammatical, even though such NCIs usually need to be licensed by an instance of sentential negation which appears to be absent in the fragment’s antecedent. Such fragments, and the contrast between () and (), have been crucial to the debate over whether NCIs should be analyzed as having inherent negative force, or rather as (a subtype of ) NPI. The same question is raised by the ambiguity of NCI fragment answers in response to a question containing negation, as in (c). The question of whether and to what extent the NPI fragment in (b) is acceptable also raises questions for the analysis of the syntax of fragments generally. These topics are taken up in section . of this chapter. Secondly, fragments can be combined with an overt negator, as in (), (), and are understood as thereby communicating the negation of the proposition that the corresponding ‘positive’ fragment communicates: ()

a. Who went to the party?—Not John (anyway). b. Where did Mary go?—Not Paris (anyway). c. Do you have any money?—Not a cent.

xorepse e Mary?— Oxi me ton John pantos. () a. Me pjon with whom danced the Mary no with the John anyway ‘Who did Mary dance with?—Not John, anyway.’ b. Who did you go to the movies with? Con Clara no. / No con Clara. with Clara   with Clara ‘Not with Clara.’

Greek

Spanish (Vicente )

If, as in one popular view, fragments are underlyingly sentential and derived by ellipsis of a full clause, then the examples in (), () are problematic, as they appear to reflect word orders which are not otherwise grammatical (e.g. *Not John went to the party). This is the topic of section .. In the remainder of this introduction, I will give some brief background concerning fragments and subsententials in general, and in particular one of the key debates in the relevant literature: given a fragment like (), is its underlying structure elliptical for the full sentence in (a) (the proposition it is understood as communicating), or is it the case that ‘what you see is what you get’ (WYSIWYG), as in (b)? ()

What did John eat?—A cake. a. [S John ate [DP a cake]] b. [DP a cake]

The first position can be termed the sententialist position (Hankamer ; Merchant ; Morgan ; Reich ; Stanley ; Weir ); the second, the non-sententialist position (Ginzburg and Sag ; Jacobson ; Stainton , , a, b, and many of the papers in Progovac et al. ). A typical non-sententialist approach, of the type exemplified particularly by Robert Stainton’s work, argues that fragments are interpreted as propositions (and ultimately as speech acts) by semantically (but not syntactically) composing them with some discourse-salient property. This property can

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



be made salient by an overt antecedent question, as in (), but could also be merely implicit in the discourse, as in (): here, non-sententialist views argue, ellipsis of syntactic structure is unlikely, as there is no linguistically available antecedent. () Q: What did John eat? → Makes salient the property ‘λx. eat(John)(x)’ A: A cake. = λP. ∃x. P(x) & cake(x) → composes to give ‘∃x. eat(John)(x) & cake(x)’ () [On getting into a taxi.] The train station, please. salient property: λx. the taxi should go to x Proponents of the sententialist view, however, point to a number of grammatical properties of fragments in support of the claim that there is more syntactic structure than meets the eye. Merchant () lists a large number of these: space allows for a summary of only one here, namely the claim that fragments appear to be dependent on A0 -movement. That is, there is evidence that the correct representation of a fragment (on the sententialist view) is not (a) (an ‘in-situ deletion’ approach), but (b) (a ‘move-and-delete’ approach): ()

a. John gave the flowers to Mary. b. The flowers John gave t to Mary.

The structure in (b) has a theoretical appeal insofar as it allows deletion to operate on a single (derived) constituent, rather than the non-continuous, non-constituent deletion that (a) would require. And there is also positive evidence for it: fragments do appear to undergo an A0 -movement step, at least in the sense that only constituents which can be A0 -moved (in a given configuration, in a given language) can be fragments. For example, the complements of a preposition in non-P-stranding languages (such as Greek, (b)) cannot be fragments: it is obligatory to include the preposition (Merchant , ’s Pstranding generalization). See Weir (: ch. ) for a list of other similar effects (but see also Ott and Struckmeier  for some important recent counterarguments). ()

a. Who did John talk to?—(To) Mary. b. Me pjon milise i Anna? with whom spoke the Anna *(Me) ton Kosta. with the Kosta

Greek

(Merchant : )

If a fragment has undergone A0 -movement, this of course implies that there is structure for it to move out from, and so data like (b) have been taken as support for the sententialist view. Move-and-delete analyses are not universally adopted even by members of the sententialist camp, however; one notable difficulty for it is that—in the relevant context, that is answering a question—the putative sentential sources (with movement) are infelicitous (i.e. English does not allow fronting of foci, only of topics): ()

What did John eat?—#A cake he ate.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

Space precludes a detailed discussion here of the relative merits of sententialist versus non-sententialist approaches, and of ‘move-and-delete’ versus ‘in-situ’ approaches within sententialism.¹ I have also not discussed the important question of whether, on an elliptical view of fragments, ellipsis is licensed under semantic identity with an antecedent, syntactic identity, or some combination. This will be addressed to some extent in what follows; for more extensive discussion I refer the reader to the overviews in Craenenbroeck and Merchant (), Merchant (), and the references cited therein. I now turn to discussion of negation in fragments, first to NCIs and NPIs.

.. N        

.................................................................................................................................. The availability of fragment NCIs as responses to positive questions, despite the fact that the negation usually required to license them appears to be absent (, ), is one of the crucial arguments in the literature for distinguishing NCIs from NPIs (see also Chapter  in this volume); this was initially noted by Zanuttini (), and the possibility of standing alone as a fragment answer is taken by (e.g.) Giannakidou (a) as a (pretheoretical) diagnostic of NCIs as a class. ()

a. ¿Qué comiste? —Nada. what eat-. n-thing ‘What did you eat?—Nothing.’

Spanish

b. *Comí nada. ate.. n-thing c. No comí  eat.. ‘I ate nothing.’ ()

nada. n.thing

a. Nani-o mita no? —Nani-mo. what- saw  what- ‘What did you see?—Nothing.’ b. *Nani-mo mita. what- saw c. Nani-mo mi-nak-atta. what- see-- ‘I saw nothing.’

Japanese

(Watanabe )

¹ It should also be noted that they are not strictly mutually exclusive approaches. Throughout the work cited above, Stainton notes that he is open to the possibility that sententialist/ellipsis accounts could be correct for answers to explicit questions, while a non-sententialist account is in his opinion better suited to account for ‘discourse-initial’ fragments, such as To the station, please on entering a taxi. See also Merchant () for discussion.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



The availability of such NCI fragments has been taken as evidence that NCIs must contain an inherent negation as part of their meaning (and so that something must ensure that sentences containing sentential negation plus one or more NCIs are interpreted as only containing a single semantic instance of negation, e.g. the “negative absorption” or “negative factorization” of Zanuttini (); Haegeman and Zanuttini (, )). The fundamental argument, as laid out by Watanabe (), is that if an elliptical/fragment answer takes a positive proposition as its antecedent, as in (), (), the fact that the answer is interpreted as negative must come from negative force within the NCI itself. Giannakidou (a, a)², who argues against an inherently negative treatment for NCIs, argues that this is not a necessary conclusion, and that the ellipsis site in () in Greek for example can be taken to contain an appropriate (sentential) negation. ()

a. Q: Ti idhes? what saw. ‘What did you see?’

Greek

A: TIPOTA. n-thing ‘Nothing.’ b. TIPOTA dhen idha n-thing  saw.

(Giannakidou a: f.)

Watanabe (), in turn, argues against this analysis, pointing out that (contentful, i.e. interpreted) sentential negation cannot freely be allowed to be inserted in ellipsis sites, at the risk of wildly overgenerating fragments with negative interpretations: (a) cannot be interpreted as (b). ()

a. Q: Nani-o mita no? what- saw  ‘What did you see?’ A: Hebi. snake b. Hebi-o mi-nak-atta snake- see-- ‘I didn’t see a snake.’

Watanabe’s argument is intuitively strong, but perhaps not watertight. It is independently known that ellipsis can (and must) sometimes tolerate some deviation from strict identity with an antecedent. The most famous example of this is Vehicle Change (Fiengo and May ), where the forms and interpretations of nouns and pronouns in an ellipsis site must be different from the antecedent in order to obviate binding theory violations or allow for possible sloppy interpretations; more pertinently to the current discussion, it appears to be possible to ‘alternate’ between NPIs in antecedents and their positive correlates under verb ² Giannakidou’s and Watanabe’s work was developed over the same period; each refers to the other.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

phrase ellipsis, as discussed in Hardt (); Johnson (); Merchant (); and originally by Sag (): ()

John didn’t talk to anybody, but Mary did talk to somebody.

This ‘flexibility’ in the content of an ellipsis site is often understood in terms of accommodation (van Craenenbroeck ; Fox ; Thoms  among many others). The basic idea pursued by Fox () is that elided material can sometimes be understood as differing from its antecedent—that is, an alternative ellipsis site is accommodated by the hearer—but only if there is some ‘trigger’ in that material which is pronounced, that prompts the hearer to carry out this accommodation. If this is generally possible, then it may be precisely the appearance of an NCI fragment—that is, a fragment which if used in a non-elliptical context would require the presence of sentential negation—that allows for the accommodation in the ellipsis site of that sentential negation, as in (). The ‘wild overgeneration’ possibility in () would not occur, not because the grammar disallows the accommodation of negation in ellipsis sites per se, but because there is no trigger (in the form of a pronounced NCI) for negation to be accommodated into the ellipsis site in (), and therefore the hearer would not perform such accommodation. However, if negation can be ‘accommodated’ into an ellipsis site due to the presence of a trigger in pronounced material, one might expect that the presence of (something which is unambiguously) a negative polarity item should also be able to trigger this accommodation. This prediction is not borne out: (b-ii) cannot be interpreted as elliptical for (b-i), even for speakers who allow NPIs to be fragment answers in general as in () (data which will be discussed in more detail below). ()

A: Did you catch any fish? a. (i) I caught some small ones, but I caught no big ones. (ii) Some small ones, but no big ones. b. (i) I caught some small ones, but I didn’t catch any big ones. (ii) *Some small ones, but/and any big ones.

()

What didn’t you catch?—Any big fish.

It is unclear, then, that negation can in fact be ‘accommodated’ into an ellipsis site, at least in fragment answers;³ so the fact that NCI fragment answers are interpreted with negative force (while unambiguous NPIs cannot be) continues to be a strong argument in favor of inherent negative force for NCIs. The data are also, however, compatible with a view that takes NCIs not to have inherent negative force—if certain assumptions are made about the formal licensing of NCIs, and how this licensing might differ from that of ‘true’ NPIs. Zeijlstra (a, a) argues that NCIs do not have the semantic force of negative quantifiers, but that they are endowed with ³ And see also Merchant (: sec. .) for discussion of the difference between VP ellipsis (e.g. examples like () and the apparent inability to change polarity between antecedent and elided clause in clausal ellipsis such as fragment answers (and sluicing).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



uninterpretable Neg features in the syntax, which need to be licensed/checked under (reverse) Agree with an instance of interpretable negation. Importantly, this interpretable negation can be provided either in the form of an overt negator, as in (), or as a covert operator, as in ().⁴ ()

a. Gianni non telefona a nessuno to n-body Gianni  calls ‘Gianni doesn’t call anybody.’ b. Gianni non[iNEG] telefona a nessuno[uNEG]

() a. Nessuno ha telefonato n-body has called ‘Nobody called.’ b. Op[iNEG] nessuno[uNEG] ha telefonato

Italian

(Zeijlstra a: ) Italian

(Zeijlstra a: )

Note that the insertion of Op[iNEG] has to be understood as something like a ‘last resort’ operation to license an instance of [uNEG] which would otherwise go unlicensed (by e.g. sentential negation), as in the subject case in (). This covert negation operator, in other words, cannot be inserted to ‘substitute’ for sentential negation, if sentential negation could otherwise be spelled out in a position where it would license the NCI; that is, cases like () are not possible. ()

*Op[iNEG] Gianni telefona a nessuno[uNEG]

Fălăus and Nicolae () propose that such an Op[iNEG] can be inserted also in elliptical contexts (as well as a few other cases I put aside here; see Fălăus and Nicolae : f.)— that is, in cases where sentential negation could not be spelled out—as a similar ‘last resort’ mechanism to license NCIs (). Note that Romanian, in contrast to Italian, would generally require sentential negation to license even subject NCIs (‘strict’ negative concord). ()

()

A: Who called? B: Nessuno. [Op[iNEG] [Nessuno [ha telefonato]]]

Italian (cf. Zeijlstra a: f.)

a. Cine a venit? — Nimeni. who has come n-body ‘Who came?—Nobody.’ [Op[iNEG] [Nimeni [a venit]]] b. Nimeni *(nu) a venit. n-body not has come

Romanian

(Fălăus and Nicolae )

⁴ This is a very schematic overview; I refer to Zeijlstra’s work for the full details.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

By contrast, ‘true’ NPIs would be taken not to have formal licensing requirements in the syntax; that is, they would not bear [uNEG] features, and the ‘last resort’ insertion of a covert negation operator would not take place.⁵ NPIs would still, however, be required to be semantically licensed (by whatever the relevant condition on NPIs is, e.g. downwardentailingness or non-veridicality), and this would not be possible in the absence of a relevant operator such as negation (either sentential negation or a covert negative operator). This is, in essence, a way of making the relevant ‘cut’ between (the licensing conditions for) NCIs and NPIs which allows for the (apparent) accommodation of negation in elliptical contexts to permit cases like (), while disallowing such putative accommodation in cases like (), partially repeated in (). () A: Did you catch any fish? B: *Any big ones. (intended: I didn’t catch any big ones) Fălăus and Nicolae () develop this analysis further in order to capture the surprising fact that, in strict negative concord languages, NCI fragment answers can convey a double negation (DN) reading if used in response to a negative question (see Chapter  in this volume), even though their putative sentential sources only allow for a single negation/ negative concord (NC) reading. ()

Cine nu a venit? —Nimeni. who not has come n-body ‘Who didn’t come?’ ‘Nobody.’

Romanian

a. You’re the first one here. [= NC: nobody came] b. Everybody’s here. [= DN: nobody didn’t come = everybody came] (Fălăus and Nicolae : ) venit. () Nimeni nu a n-body not has come ‘Nobody came.’

Romanian [only NC, 6¼ ‘everybody came’] (Fălăus and Nicolae : )

If we suppose, with Fălăus and Nicolae (), that the insertion of a covert negation operator is only possible in cases where sentential negation is unavailable (e.g. in elliptical cases), that rules out a double negation reading for a case like (), but allows for covert negation to be inserted—optionally⁶—in the elliptical context in (). Importantly, this ⁵ A reviewer asks whether, on this view, NPIs have any polarity features or other features in common with NCIs. As far as I understand, there would not need to be, and perhaps could not be, any formal (syntactic) polarity feature in common between the two. How to reconcile this with the discussion in Merchant (), in which NPIs (or rather existential quantifiers) crucially do have a polarity feature which is valued under agreement with sentential polarity (an underspecification analysis which allows for the kind of alternation under ellipsis seen in example ()), is a tension I have to leave aside here. ⁶ A reviewer raises concerns about the extent to which this optionality is constrained. The process would have to be constrained by (i) only being available in elliptical contexts, and (ii) only being available if an NCI requires licensing (‘last resort’). Some analog of (ii) would be required in any

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



covert negation introduces a negative force in addition to the sentential negation contained within the ellipsis site, leading to the availability of a double negation reading. (I omit here some details concerning the scope position of the NCI with respect to the two negations; see Fălăus and Nicolae : .) ()

a. [Nimeni [nu a venit]] b. [Opi [nimeni [nu a venit]]

(= single negation, ¬∃x P(x)) (= double negation, ¬∃x ¬P(x) = 8x P(x))

An alternative account for these data may be to say that covert negation is always inserted alongside a fragment NCI (to formally license it), and the distinction in () concerns optionality in whether the sentential negation is taken to be present in the ellipsis site.⁷ This is also presumably the tack that those who analyze NCIs as inherently negative (and who are unwilling to countenance covert negation) would need to take in order to capture the double negation reading in examples like (). Another possible alternative would appeal to lexical ambiguity in NCIs, along the lines of Herburger (): Espinal and Tubau (a) report that some speakers of Spanish and Catalan (non-strict negative concord languages) also allow for double negation readings for NCI fragments with negative antecedents, which they analyze as lexical ambiguity in NCIs between polarity items and negative quantifiers. Fragments, then, while of crucial importance to the debate, do not necessarily offer watertight arguments for or against any particular view of NCIs. We can note in passing, however, that non-sententialist analyses of fragments (at least if adopted strictly) seem forced to assume that NCIs have (or can have) an inherently negative reading: if NCIs are simply indefinites, and if fragments are simply words or phrases in isolation, then NCI fragments should either have a (positive) indefinite reading or simply be ungrammatical (as they have no licensor). A non-sententialist account which took the view that NCIs did not have inherent negative force would at minimum require a covert constituent negation in

approach to NCIs which countenances a covert negation approach to their licensing, but it is clear here that a lot rests on the precise formulation of ‘last resort’. See Breitbarth (: ), Espinal and Tubau (a), and Zeijlstra (a: ) for some discussion. ⁷ Note also that, in English, a double negation reading is possible for (i), and—to my ear—a single negation reading is also available, though more marginally: (i) Who didn’t come?—Not the students (anyway). a. [not [the students [didn’t come]]] = the students didn’t not come = the students came b. ?[not [the students [came]]] = the students didn’t come If that’s right, then it may indeed be possible to ‘shift’ from a negative antecedent to a positive ellipsis site, as long as negation is overtly expressed in some way—either by a negative particle as in English, or by the presence of an NCI in negative concord languages. Cf. the discussion of negated fragments in section .. As a reviewer points out, the behavior of polarity responses such as yes and no, and the ambiguity (or intermediate acceptability) of responses to negative questions such as (ii), are surely also relevant here; space precludes detailed discussion here, but see Brasoveanu, Farkas, and Roelofsen (); Holmberg (); Kramer and Rawlins (); Roelofsen and Farkas () for relevant discussion. (ii) Did John not write to Mary?—?Yes. a. ?= John did not write to Mary. b. ?= John did write to Mary.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

construction with an NCI fragment, as in (), which would be possible as such but which would arguably go against the WYSIWYG spirit of such analyses (and cf. the upcoming discussion in section . concerning the feasibility of such constituent negation analyses in general). () a. [[nessuno]] = λP. ∃x. P(x) & person(x) b. [[Neg]] = λX. λYσ. ¬X(Y) c. [[ [Neg nessuno] ]] = λP. ¬∃x. P(x) & person(x) A last point to consider in this section is the status of NPI fragment answers. As noted above, these are ungrammatical in response to positive questions, but at least some English speakers accept them in answer to negative questions. There appears to be some interspeaker variation on this point—and see Temmerman () for discussion of the (un)acceptability of similar NPI fragments in Dutch—but all of den Dikken, Meinunger, and Wilder (), Valmala (), and Weir () report examples like () to be grammatical. () A: (I know what John did bring to the party, but) what didn’t he bring? B: %Any wine. (After den Dikken, Meinunger, and Wilder : ) Valmala () and Weir () argue that the availability of such fragments argues against a move-and-delete analysis of fragments; given the movement in (), the NPI any wine is outside the scope of its licensing negation. ()

[Any wine [he didn’t bring t]]

Valmala () uses this fact (among others) to argue in favor of a kind of nonsententialist analysis of fragments. For den Dikken, Meinunger, and Wilder (), it is evidence in favor of an in-situ sententialist account without movement. Weir () argues that fragments do move, but only at PF in order to escape an ellipsis site, not at LF; the NPI fragment in a case like () would then be within the scope of its licensor at LF. Any of these approaches can in principle account for the datum in (). That datum, however, may not necessarily be problematic for a movement account if NPIs can (in certain circumstances) be permitted to reconstruct to a position under their licensing negation, as in the famous example in (a)⁸ (see de Swart (), as well as Uribe-Echevarria () and other references cited therein), which contrasts with ungrammatical cases such as (b). ()

a. A doctor with any reputation was not available. b. *Anyone was not available.

I refer to de Swart () for a pragmatic analysis of why inverse scope (reconstruction) should be possible for the NPI in cases like (a) but not (b); importantly, a contrast ⁸ Sauerland and Elbourne () discuss similar cases and argue that they arise because of movement of the subject at PF only, the motivation for Weir (: ch. )’s PF-movement account of similar fragments.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



similar to () appears to obtain with NPI fragments, potentially accounting for some of the divergent speaker judgements (Merchant (: ) marks examples similar to (b) as ungrammatical, for example). ()

Who were you not able to consult? a. A doctor with any reputation. b. ??Anyone.

This difference would then be consistent with NPI fragments having fronted as in (), with inverse scope (i.e. reconstruction) being licensed in () and (a) but not in (b).

.. N 

.................................................................................................................................. I now move to discussion of fragments which co-occur with a negator, as in (). ()

Who came to the party?—Not Ben, anyway.

Merchant (, ) compiles data from a number of languages for the closely related structure of “negative stripping”, (), (). I will take such cases of stripping to be derived by the same process as fragment answers (see Merchant () for justification for this move). ()

Anna Anna Anna Anna Anna

()

I Anna Anna Anna Anna

left ging ging gikk est partie

efige è partita poshla left

but aber maar en mais

not nicht niet ekki pas

Ben. Ben. Ben. Ben. Ben.

alla oxi o Ben. ma Ben no no Ben njet but no Ben no

English German Dutch Icelandic French (Merchant : ) Greek Italian Russian (Merchant : )

There is cross-linguistic variation along a number of axes concerning such structures, as Merchant discusses. First, languages differ in whether the negator in such structures is the standard sentential negator (parallel to English not), or a particle that could be used as a standalone response (that is, parallel to English no).⁹ Merchant identifies the crosslinguistic difference as being between those which use phrasal negation (e.g. English, German, Dutch, French), which use this phrasal adverb in negative fragments, and those ⁹ As Merchant notes, some languages use the same word for both purposes (e.g. Spanish no), and so it is impossible to tell which category they belong to.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

with head or affixal negation, such as Greek and Italian (head) and Swahili and Turkish (affixal), which use the response particle. Merchant notes that head negation never appears to be used in such structures; languages which express negation as a head or a verbal affix but which lack a response particle ‘no’, such as Irish, simply lack structures like (). A second axis of difference concerns the linear order of the fragment and negator. The ‘adverbial’ languages seem to systematically place the negator to the left of the fragment, as shown in (), but the ‘particle’ languages vary; for example Greek places the negator to the left, but Russian places it to the right. In Spanish, both orders are possible, as Vicente () discusses: ()

Who did you go to the movies with? a. Con Clara no. b. No con Clara. Spanish with Clara   with Clara ‘Not with Clara.’ (Repeated from (b), Vicente : )

Vicente notes that this difference in order gives rise to a difference in interpretation: (a) is the unmarked form, while (b) gives rise to a corrective interpretation, that is it repudiates a previous (possibly implicit) assertion that the speaker did go to the movies with Clara, and also presupposes that the speaker did indeed go to the movies with someone (just not with Clara). See also Servidio () for discussion of similar facts in Italian. As Merchant () notes, there are in principle two analyses that can be adopted to deal with negated fragments of this type. The first is to assume that not forms a constituent with the fragment, an analysis proposed by Depiante () for Spanish. This analysis is the only one available to nonsententialist/WYSIWYG approaches, shown in (a); it is also compatible with in-situ sententialist analyses, as in (b), and move-and-delete analyses, as in (c). ()

What did John give to Mary?—Not flowers (anyway). a. [not [DP flowers]] b. John gave [not [DP flowers]] to Mary c. [ [not [DP flowers]]i [John gave ti to Mary]]

In the second possible analysis, compatible with a sententialist analysis but not a nonsententialist one, negation does not form a constituent with the fragment, but rather inhabits a left-peripheral position above the deleted clause, shown for an in-situ analysis in (a) and a move-and-delete analysis in (b). I will refer to this as a left-peripheral negation analysis. () a. [Not [John gave flowers to Mary] b. [Not [flowersi [John gave ti to Mary]]] Neither of these analyses are immediately highly satisfactory, particularly for sententialist analyses, which seem forced to assume underlying word orders that aren’t grammatical. () a. b. c. d.

*John gave not flowers to Mary. *Not flowers John gave to Mary. ??Not flowers did John give to Mary. **Not John gave flowers to Mary.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



The constituent negation analysis, however, may on the face of it be more initially promising given that (c) is perhaps less bad than the others (especially the left-peripheral negation in (d)), that (a) becomes good in construction with but (), and that constituent negation is grammatical in a range of other contexts (); in fact, possibly it is only ungrammatical in cases like (a), where it appears to be ‘pre-empted’ by sentential negation, John didn’t give flowers to Mary. ()

John gave not flowers but chocolates to Mary.

()

a. b. c. d.

He found something interesting there not long ago. He had spoken with someone else not many hours earlier. He married a not unattractive girl. Writers not infrequently reject suggestions. (Klima : , taken from Toosarvandani : )

There are, however, problems for the constituent negation analysis—or at least indications that it is not the only available strategy. One problem comes from interpretation, as Merchant () notes in cases like () (see Vicente (: ) for equivalent Spanish data). () A: What did Beth say she wanted to study? B: Not French. The fragment in () can have the reading ‘Beth didn’t say she wanted to study French’, that is that there was no event of Beth saying this. (It is compatible with Beth wanting to study French but happening not to give voice to this opinion, for example.) However, if only constituent negation were available, () could only mean something like ‘Beth said she wanted to study not French’ (i.e. Beth did say something, and that something had the content that she did not want to study French): there would be no way of getting constituent negation to scope over said in the requisite way. There are also purely syntactic problems with a constituent negation analysis. For example, such an analysis does not immediately have anything to say about the Spanish cases in (a), Con Clara no ‘with Clara ’, where the negator comes after the fragment: this is not a possible word order for constituent negation in Spanish. In addition, Merchant (, ) notes other cases which appear to involve clausal ellipsis and not (). In these cases, a constituent negation analysis seems untenable, as there is no constituent to negate. () a. If he comes, it’ll be fine; if not, we have a problem. (=if he does not come . . . ) b. I don’t know whether he’ll be there or not. (=or whether he will not be there) To which we can add: ()

Is he coming?—Possibly not.

In (), not is not a constituent negation of possibly (which would give the wrong scope); rather, not has to be taken to be negating the (elided) sentence he’s coming (with possibly taking widest scope over the negated sentence).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

It should be noted that all of these data and arguments are compatible with constituent negation being one available strategy for negated fragments (as Vicente :  notes). But it does appear that a sentential negation parse has to be available in addition.¹⁰ In-situ sententialist analyses make the right predictions with regard to word order in a number of cases: () a. What did John give to Mary?—Not flowers, anyway. = John did not give flowers to Mary anyway b. Is he coming?—Possibly not today. = Possibly he is not coming today But not with respect to subjects: ()

Who danced with Mary?—Not John, anyway./*John not, anyway. 6¼ John did not dance with Mary anyway.

And it is not immediately obvious why head negation (either alone, as with Greek dhen in (b), or affixed onto a verb) would not be a grammatical remnant under such in-situ deletion. (This is a particular case of the general problem faced by in-situ accounts that heads¹¹ cannot be ellipsis remnants.) ()

a. Who did Mary dance with?—*Didn’t John, anyway. 6¼ Mary didn’t dance with John anyway b. {Oxi/ *Dhen} me ton John pantos. no not with the John anyway

Greek

Merchant () and Vicente () argue that negation in these contexts is hosted in a left-peripheral NegP (or ΣP/PolP), above the Focus position to which fragments are assumed (on a move-and-delete analysis) to move in English. Vicente in addition assumes that (at least in Spanish) a further projection, TopP, is available to host movement— thereby allowing for the two orders in () (and, to the extent that information-structural properties can be ‘read off ’ the cartographic labels TopP and FocP, also accounting for or at least encoding the information structure differences between the two answers). If ΣP can also host positive polarity, then this structure can also account for the positive fragment answer in ().¹²

¹⁰ To the extent that nonsententialist analyses of fragments can only appeal to constituent negation, this can then be taken as an argument that a sententialist parse must at least be one of the available structures for fragments. ¹¹ That is, non-phrases: single-word constituents that are simultaneously heads and phrases (e.g. intransitive verbs/VPs in cases like What will he do then?—Dance) can be fragments. ¹² But see Servidio () for a slightly different account in which the polarity particles occupy FocP itself, following Holmberg (, ). On ellipsis accounts of polarity particles in general (a matter which is clearly relevant but which space unfortunately precludes a detailed discussion of ), see references in footnote .

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   ()



[TopP [ΣP not [FocP Johni [TP . . . ti . . . ]]]]

() a. No con Clara. ‘not with Clara’ [ΣP no [FocP con Clarai [TP . . . ti . . . ]]] b. Con Clara no. ‘with Clara not’ [TopP con Clarai [ΣP no [TP . . . ti . . . ]]]] ()

Con Clara sí. ‘with Clara yes’

While the idea that polarity can be expressed in a very high position in the clause is not a new proposal (dating back at least to Laka (), and see also Chapter  in this volume), the analysis in () does still have to reckon with the fact that such negation does not generally appear in this position in English or Spanish outside of elliptical contexts, a restriction which both Merchant and Vicente simply have to stipulate. In addition, it is not immediately obvious why phrasal negation is required, that is, why head negation apparently cannot be inserted in Σ⁰ in ().¹³ In sum, then, constituent negation analyses of negated fragments may be correct for a subset of cases but cannot account for them all; but sentential analyses face the problem of otherwise unattested word order (whether they assume negation is in-situ, or in a leftperipheral position). Space precludes a detailed discussion of which view is correct here, but the left-peripheral account is sufficiently attractive especially given the data in () that I would like to briefly sketch two possible views of how its problems might be resolved. One view would take a lead from the discussion of the insertion of covert negation in section ., as proposed by Fălăus and Nicolae (). Recall that they proposed that negation is generally insertable in a left-peripheral position (following Zeijlstra a, a), but only in elliptical contexts, and only as a ‘last resort’: the intuition being that “you cannot appeal to covert negation if you have the space where you could have spelled out an overt negation” (i.e. ‘normal’ sentential negation, Fălăus and Nicolae : ). In the cases they were concerned with, the negation was covert, and the ‘last resort’ was to license an otherwise unlicensed negative concord item. Suppose that in negative fragment cases like Not John, negation can also be inserted in a left-peripheral position under ellipsis, as a ‘last resort’ in order to indicate the negative polarity of the response: while such negation can be covert in construction with an NCI (because negation is overtly signaled by the morphology on the NCI itself), it must be overt if the fragment does not bear any negative morphology, as in the examples under discussion here.¹⁴ This proposal would tie

¹³ And indeed, in those cases other than ellipsis where a left-peripheral negation phrase has been adduced for (dialectal varieties of ) English (e.g. Foreman ; Green )—cases of apparent negative auxiliary inversion such as Ain’t nobody take the bus—the negator appears to be a head (an n’t-marked verb), not an adverb. ¹⁴ Vicente (: ) notes that overt negation is in fact incompatible with an NCI fragment (i), which is compatible with the view sketched here. (i)

A: What has John read? B: (*No) ningún libro. not n-one book

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

together the syntax of negation independently proposed for negative concord languages with the syntax of negative fragments in non-negative concord languages like (Standard) English—though it would await elaboration of the precise nature of the ‘last resort’ condition on the insertion of left-peripheral negation. It is also unclear whether it necessarily captures the ‘adverb/phrasal negation only’ restriction discussed in Merchant (, ). This could potentially hinge on the question of the nature of the leftperipheral negative operator and whether head negation is ever taken to independently contribute (semantic) negative force in the absence of such an operator; see Zeijlstra (a, a) for discussion. An alternative is to suggest that the left-peripheral not is in fact normal sentential negation which has moved from its middlefield position to a left-peripheral position. ()

[Noti [flowersj [John ti gave tj to Mary]]]

This would then be a case of exceptional movement, that is, a kind of movement that is only triggered in elliptical situations. Such movement has been appealed to in the analysis of other elliptical phenomena; for example pseudogapping (a) is often analyzed as movement of an object out of the (VP) ellipsis site ((b); Fox and Pesetsky ; Gengel ; Jayaseelan , ; Johnson ; among others). As (c) shows, such movement is ungrammatical absent ellipsis. () a. He would eat more potatoes than he would carrots. b. . . . than he would [XP carrotsi [VP eat ti]] c. * . . . than he would carrots eat. And in fact the kind of movement appealed to in the move-and-delete analysis of fragments is ‘exceptional’ in general in English (see Boone a, b; Weir ): the putative sentential source for () on the move-and-delete analysis is infelicitous. ()

What did John eat?—#Chipsi he ate ti.

If we grant that movement is indeed ‘exceptionally’ possible out of an ellipsis site, then we might contemplate that not (and similar negative adverbs in other languages) can also undergo such movement, as in (). If such movement is A0 -movement—which can apply only to phrases, not heads—then we would also understand why negation in fragments is restricted to adverbial/phrasal negation. It may also give us a handle on other apparently out-of-place adverbs in fragments, such as (). ()

A: Who wins the pub quiz each week? B: Never our team (anyway). B0 : Our team—always our team.

()

a. Our team {never/always} wins. b. *Never our team wins. c. *Always our team wins.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   ()



Putative exceptional movement of the adverb under ellipsis: [ always/neveri [ our teamj [tj ti wins]]]

While an ‘exceptional movement’ analysis has these advantages, it should be noted that it still needs to stipulate the landing site for ‘exceptionally moved’ negation (to rule out the word order *John not). It is also not immediately compatible with the ‘double negation’ facts mentioned in footnote , where the putative left-peripheral negation would have to be additional to the sentential negation and so the former could not be a moved instance of the latter. The correct syntax for negated fragments, then, is a project which still awaits a complete answer.

.. C

.................................................................................................................................. This overview has only scratched the surface of the interaction of negation and fragments, but it is clear that further research in each of these domains has the potential to feed the other. Negation clearly provides one way, for example, to potentially adjudicate between different theories of ellipsis identity and the sententialist/non-sententialist dispute: as argued above, in the absence of an elaboration of the possible interpretations of constituent negation, the availability of negative fragment answers appears to speak in favor of a sententialist approach (or at least suggests that a sententialist analysis is available). The interpretation of fragment NCIs, especially the availability of single- and double-negation readings, also provides a testing ground for the question of how ‘flexible’ the identity relation between antecedent and ellipsis site can be (e.g. whether it is possible to ‘accommodate’ a different polarity in an ellipsis site from the antecedent). Conversely, further research on identity and accommodation in ellipsis can itself shed further light on the question of whether NCIs must be taken to bear an inherent negative semantics; for example if we find incontrovertible evidence that the content of ellipsis sites cannot be accommodated in the way that ‘NPI’ theories of NCIs would require. As an anonymous reviewer notes (see also footnote ), a concrete avenue for further empirical research here would be the interaction of polarity particles (yes and no) with questions containing NCIs. In sum, many interesting problems remain in this domain—both for the negation theorist and the ellipsis/fragments researcher.

A

.................................................................................................................................. I thank two anonymous reviewers for their comments, as well as Yiannis Kokosalakis for Greek examples presented without other attribution. Any errors are of course mine.

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi

  ......................................................................................................................

         ......................................................................................................................

 

.. I   

.................................................................................................................................. We talk about ‘negative concord’ (NC) when we have a single interpretation of negation in the face of multiple apparent negative exponents. Observe: visto niente. () a. Gianni *(non) ha Gianni  have. seen n-thing ‘Gianni hasn’t seen anything.’ b. *(No) he dit res.  have. said n-thing ‘I didn’t say anything.’ c. He didn’t say nothing.

Italian

Catalan

English: Double negation

As can be seen, one occurrence of negation is the negative marker, and a second occurrence is the so-called ‘n-word’. Laka () used the term because of the presence of the negative marking n. We will use the term ‘negative concord items’ (NCIs) to refer to this class. Other terms have been used for negative concord too: “double attraction,” “neg-incorporation” (Klima ), “negative attraction” (Labov a). In the face of examples such as above, we say that Italian and Catalan are ‘negative concord languages’; but English is not because the combination of negation n’t and nothing results in double negation. The English equivalent of the NC structures in (a,b) involves negation and the negative polarity item (NPI) any, which lacks negative meaning, as we see in the translations.¹ ¹ There are varieties of English that exhibit negative concord; Labov (a) was the first to illustrate extensive patterns of concord in African American Vernacular, and for more recent discussion see Green (). Concord is also attested in Southern dialects of English in the US (Edmiston ).

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi

     



It becomes immediately obvious that negative concord and NPI phenomena are related: in concord, the NCI requires the presence of negation, pretty much the way any does. Here are some more examples: () a. Balász *(nem) látottt semmit. Balász  saw. n-thing ‘Balász didn’t see anything.’ b. Milan *(ne) vidi nista. Milan  see. n-thing ‘Milan doesn’t see anything.’ c. *(Dhen) ipa TIPOTA.  said. n-thing d. John-wa nani-mo tabe-*(nak)-atta John. n-thing eat.not.past ‘John didn’t eat anything.’

Hungarian

Serbian/Croatian

Greek Japanese

French, Hungarian, Serbian/Croatian, Greek, and Japanese are ‘negative concord languages’; without negation, as can be seen, the NCIs are ungrammatical just as any would be. Some NCIs have emphatic accent (as indicated above for Greek, see Giannakidou , a; Giannakidou and Yoon ; Chatzikonstantinou  gives precise phonetic characteristics of the accent.). Unlike any, the NCIs appear to contribute negative meaning ‘in isolation’, as Zannutini () noticed, for instance, in fragment answers: () Q: Who did you see by the window? A: Nobody/Nessuno/Nadie/Ningu/Niemand/*Anybody. The grammatical fragments observed in (A) are NCIs; anybody, as we see, is ungrammatical. NCIs appear therefore equivalent to negative quantifiers in fragment answers. Giannakidou (a) offers the following definition of NCI: ()

An expression α is an NCI (aka ‘n-word’) iff: (i) α can be used in structures containing sentential negation or another α-expression, yielding a reading equivalent to one logical negation; and (ii) α can provide a negative fragment answer (i.e. without overt negation).

Under this definition, Germanic words such as nobody, niemand, are not NCIs since when they co-occur with negation they produce a double negation (DN) reading: ()

a. John didn’t see nobody. DN: It is not the case that John didn’t see anyone. (Entails: John saw someone).

niet niets gezegd. () Ik heb I have.  nothing said ‘I haven’t said nothing’ = ‘I have said something’

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi



 

The default double negation reading is evidence of negativity. Unlike negative quantifiers, any is not semantically negative, hence we do not talk about NC in this case. Germanic negative quantifiers are n-marked, but they are not NCIs. Crucially, Giannakidou’s definition of NCIs in (), correctly, does not appeal to morphological negative marking. The definition also does not presuppose that NCIs are semantically negative. Some NCIs are morphologically marked with n- (ningu, nessuno) but others are not (e.g. Catalan res, Greek and Japanese NCIs); some NCIs are wh-marked (e.g. in Japanese and Korean); and some have emphatic accent, as we said. From the perspective of negative form, then, NCIs are not marked negative as a class. If we assume, based on the fragment answer criterion, that NCIs are semantically negative, we cannot explain why two or more NCI occurrences still produce concord: nikoho. *(ne-)volá () Dnes nikdo Today n-person -call. n-person ‘Today nobody calls anybody.’

Czech

TIPOTA. () KANENAS *(dhen) ipe n-person  said. n-thing ‘Nobody said anything.’

Greek

semmit. *(nem) látott () Senki n-person  saw. n-thing ‘No one say anything.’

Hungarian

Here we have two occurrences of NCIs plus negation, but semantically a single negation. The concord reading, thus, seems to create a paradox: if NCIs are negative (‘concord’), in order to get the concord reading we must admit that somehow they can lose their negative force; but if we assume that NCIs can lose their negative force, then they cannot be inherently negative and we can no longer talk about concord, certainly not in all cases. As I said at the beginning, it seems to be a robust generalization that some languages have negative concord and others do not. This suggests that negative concord must be some kind of rule, and that languages parametrize as to whether they have it. But why would a language have such a rule? It seems redundant. Here it is important to remember that natural language is full of redundancy, known as agreement, where we have multiple exponents of what appears to be a single semantic category (e.g. gender, person, number, tense, etc.). It is also no accident that many languages that have negative concord also tend to have other kinds of agreement. Naturally, agreement has been invoked early in the literature (Haegeman and Zanuttini , ; Zanuttini ; Zeijlstra a; Haegeman and Lohndal ), producing a non-violable grammatical requirement that explains also the clear judgment in the sentences discussed so far, namely that without negation the NCIs are ungrammatical. Crucially, the agreement analyses do not have to assume that NCIs are semantically negative; agreement can be stated at the morpho-syntactic level while also arguing that semantically NCIs are not negative (as is done e.g. in Zeijlstra’s a and subsequent work).

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi

     



In negative concord, NCIs behave like negative polarity items (NPIs): they need negation to be ‘licensed’ (Giannakidou , , a, a; Giannakidou and Zeijlstra ), hence in these languages ‘negative’ agreement and NPI-licensing would appear to be syntactically related phenomena. And since the NPI any is not negative, a reasonable idea is that NCIs are also not negative. This has indeed been a productive line of research going back to Ladusaw (), Acquaviva (), Giannakidou (), Giannakidou and Quer (), and Déprez (, ), to mention some of the earlier works in the s. In understanding the phenomenon of negative concord we must be able to appreciate both the contribution of negation and the contribution of NCIs, and we should be able to acknowledge certain key aspects of variation (see some recent discussion in Giannakidou a, Déprez , and Giannakidou and Zeijlstra ). The strict negative concord (NC) variety, that is the variety we observed thus far requiring sentential negation, is in fact strikingly uniform across the languages that exhibit it. Strict NC seems to present a synchronically and diachronically stable pattern (Chatzopoulou )—in contrast to non-strict NC, which appears more volatile. More diverse patterns are observed in Romance languages and Creoles derived from Romance, as well as West Flemish and Afrikaans. Romance Creoles can be strict NC too (see Déprez and Henri ). I will not offer extensive historical surveys in this chapter; for such see Giannakidou (a), Giannakidou and Zeijlstra (), also de Swart, this volume, Espinal and Prieto this volume. My main goal here is to entertain hypotheses at a level of generality useful enough to guide new researchers. As I just said, in strict NC, it is easier to give an answer and advance the argument that NCIs are not negative and NPI like. In the landscape of Romance NCIs, however, the situation is more complicated, and the most plausible hypothesis is that there are different kinds of Romance NCIs: some negative, some not negative, and some ambiguous. Ambiguity approaches include Déprez (), Déprez and Martineau (), Herburger (), Espinal and Tubau (a), Martins (), to mention just a few. Routine double negation must be taken as baseline evidence for inherent negativity in the NCI, just as it is in the case of negative quantifiers. In strict NC, double negation readings with NCIs do not arise by default (for Greek, see Veloudis , Giannakidou , a). The literature does mention DN readings in Romanian (Fălăuş ; Iorda˘chioaia ), and Hungarian (Puskás ); these readings, however, are marked with distinct intonation, and do not routinely arise. In non-strict NC languages, likewise, such as French Catalan, or Spanish (de Swart and Sag ; de Swart ), DN are reported to be possible, but they are never default (Déprez et al. ; Espinal and Tubau (a); Espinal et al. ). Double negation seems to be attested in Afrikaans or West Flemish (Biberauer and Zeijlstra ). In recent work, Déprez et al. (), Espinal and Tubau (a), and Espinal et al. () provide experimental confirmation that NC in Catalan is the preferred and faster interpretation for negative sentences. The discussion proceeds as follows. In section ., I introduce the strict vs. non-strict negative concord distinction, which is instrumental in allowing us to distinguish between stable and less stable patterns of concord. We also discuss in this section the question of what drives NC: the negation or the NCI. In section . we discuss the hypothesis that NCIs are indefinites, in section . the idea that they are universal quantifiers, and in section . the negative approach.

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi



 

.. S  -  

.................................................................................................................................. Giannakidou () distinguishes two varieties of negative concord: strict and non-strict. Strict NC shows little variation across languages, and is almost identical to an NPI dependency. Non-strict negative concord, on the other hand, allows for the omission of sentential negation if the NCI is preverbal, and characterizes Romance NC, and to a lesser extent West Flemish and Afrikaans.

... Strict NC, negation, and change Giannakidou () defines as ‘strict’ the variety of NC where the marker of sentential negation must be present with NCIs at all times, and regardless of their number. Recall the Czech, Greek, and Hungarian examples mentioned earlier: ()

KANENAS *(dhen) ipe TIPOTA. n-person  said. n-thing ‘Nobody said anything.’

()

semmit. Senki *(nem) látott n-person  saw. n-thing ‘No one say anything.’

Greek

Hungarian

The negative marker cannot be omitted despite the fact that we have two NCIs. Strict NC is quite common. Many language families, often unrelated, have it: Slavic languages, Greek, Albanian, Hungarian, Japanese, and Korean are strict NC languages. The sentential negation marker is the logical vehicle of negation, and it is obligatory with any number of NCIs: ()

KANENAS *(dhen) ipe TIPOTA POTE se KANENAN. n-person  said. n-thing n-ever to n-person ‘Nobody has ever said anything to anyone.’

Greek

Strict NC seems to be the easiest to handle as an NPI dependency: NCIs are NPIs (Giannakidou , a, a; Giannakidou and Zeijlstra ), and are as negative in strict NC as any is. One important property of strict NC is that it seems to be negation centered. Strict NC presents a prime case for what Déprez () and others would characterize as a macroparametric approach that concentrates on sentential negation as the key factor in negative concord relations. (Early approaches based on the Neg Criterion, which sought parallels between negative relations and wh-dependencies, were likewise also macro-parametric.) Micro-parametric approaches, on the other hand, define patterns “not over negative

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi

     



concord E-languages, but over types of internal-structures independently of their actual proportional embodiment in particular languages, dialects, sociolects or historical stages” (Déprez : ). The two patterns are opposite: what ‘drives’ NC in the one approach is the negation, and in the other is the NCI. However, if we take the parallel with NPIs faithfully, we must accept that NPI licensing is ‘driven’ equally by the licenser and the licensee: it is the properties of any that make it sensitive to negation (and other nonveridical operators, Giannakidou  et seq., Zwarts ). The licensing dependency goes both ways, and it is perhaps not even interesting to establish what comes first: the sensitivity of an NPI or the non-veridicality of the licenser. The question of what comes first (what ‘drives’) is more meaningful in the context of historical change, and here the discussions center around the so-called Jespersen Cycle. The Jespersen Cycle is a generalization that seeks to explain how negation systems change, and postulates, in its classical version, that it is phonological weakening of sentential negation that drives the change. Regarding Greek, Kiparsky and Condoravdi (), and especially Chatzopoulou (), argue for a refinement of the traditional understanding, in a way that abstracts away from morpho-syntactic and phonological particulars, and explicitly places its regularities in the semantics. Chatzopoulou argues that the Jespersen Cycle is a phenomenon which targets intensified predicate negation, and with time, transforms it into propositional negation (Chatzopoulou :  ()). The Jespersen Cycle includes two main stages: one in which an element functions as intensified predicate negation and another in which the same element becomes sentential (propositional) negation. A number of microstages can be involved and this change is represented as upward lexical micro-movement. Crucially, the phenomenon is negation driven: it is the semantic change in the negation that produces the patterns of change, Chatzopoulou argues (and offers evidence for Greek covering a ,-year period).

... Non-strict negative concord Giannakidou () defines as non-strict the patterns of negative concord that do not necessitate the presence of sentential negation. Non-strict patterns exhibit considerable variation—and in at least some usages NCIs appear to have negative meaning. In non-strict concord, an NCI can appear in preverbal position with negative meaning without negation; a preverbal NCI can also license a postverbal one, again without negation. Here are examples from Italian: ()

Ieri nessuno ha telefonato. Yesterday n-person has called ‘Yesterday nobody called.’

()

a. *Gianni ha telefonato nessuno. John has called n-person b. Ieri *(non) ha telefonato nessuno. Yesterday  has called n-person ‘Yesterday nobody called.’

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi

 ()

  Ieri nessuno ha parlato con nessuno. Yesterday n-person.body has talked with n-person ‘Yesterday nobody talked with anybody.’

Preverbal nessuno appears to be the negation in the sentence, yet postverbally it needs negation. The example in () with multiple NCIs, one preverbal and without negation, illustrates negative spread (after den Besten ). These two properties, that is the ability of NCIs to appear preverbally without negation and negative spread, are characteristic of nonstrict NC patterns. If indeed the NCI preverbally is negative, how come it needs sentential negation in the postverbal position? This conflicting behavior has puzzled theorists for decades. And to make things trickier, there is variation with respect to the presence of the negative marker or not. Some languages forbid it; others allow it. We can think of the first class, as Giannakidou and Zeijlstra () put it, as “strictly Non-strict” NCIs. In this variety— predominantly attested in Italian, Spanish, and Portuguese—postverbal NCIs must be accompanied by a preverbal negative element (an NCI or a negative marker), but preverbal ones must not be followed by the negative marker. To continue with Italian, we see below that if we add the negative marker non, a preverbal NCI produces double negation, as in Germanic: ()

a. Ieri nessuno non ha telefonato. Yesterday n-person  has called ‘It is not the case that yesterday nobody called.’

Italian

b. Ieri non ha telefonato nessuno. Yesterday  has called n-person ‘Yesterday nobody called.’ With preverbal NCI and negation, the only possible reading is the double negative. This suggests that the NCI in the preverbal position must be interpreted negatively. The second type of non-strict NC can be called “optionally Non-strict” (Giannakidou and Zeijlstra ). In West Flemish, for instance, the negative marker with NC is possible with preverbal NCIs, without a double negation reading (cf. Haegeman ; Haegeman and Zanuttini ); likewise in certain varieties of Catalan (Quer ; Vallduví ): ()

. . . da niemand Valère (nie) ken. . . . that n –person Valère  knows ‘ . . . that nobody knows Valère.’

()

Cap estudiant (no) va dir res. n-student  aux say n-thing ‘No student will say anything.’

West Flemish

Catalan; restricted varieties

Catalan thus differs from Italian (and Spanish, Portuguese) in allowing negation with the preverbal cap without a double negative reading. Hence cap seems to give rise to a truly ambiguous construal. In postverbal position, Catalan NCIs like Italian need negation, and

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi

     



have been claimed to exhibit a polar behavior (Linebarger ; Progovac ; Giannakidou , a; Martins ). Déprez et al. () note that NCIs are sensitive to the nonveridical property of a c-commanding licenser (examples below from Déprez et al.): ()

No ha comprat cap dels llibres.  has bought none of.the books ‘(S)he has not bought any of the books.’

()

Ha comprat cap dels llibres? has bought any of.the books ‘Did she buy any books?’

()

Si ha comprat cap dels llibres, jo ho hauria de saber. if has bought any of.thePL books I it should of know ‘If (s)he has bought any of the books, I should know it.’

In the non-negative polarity contexts, Catalan NCIs are unquestionably non-negative existentials, NPI-like. Nevertheless in the preverbal position the NCI can appear optionally without negation. Finally, French is an NC language that has a negative marker pas and an additional expletive negative marker ne. Although pas is the sentential negator, it is systematically banned from negative concord (see Corblin et al. ; Penka ; Zeijlstra b). A sentence with one or more NCIs and pas always gives rise to DN reading (Déprez ; de Swart and Sag ; de Swart ): ()

Personne (ne) mange pas rien. N-person  eats  n-thing ‘Nobody eats nothing.’ = ‘Everybody eats something.’

French

On the other end in Romance, we have Romanian which is reported to be strict concord. The data below are from Giannakidou (a) who cites Bernini and Ramat (: , ): ()

a. Nimeni *(nu) vine. n-person  come. ‘Noone is coming.’ b. Nimeni *(nu) vazu nimic. n-person  saw. n-thing ‘Nobody saw anything.’

The typological picture of NCIs with respect to the position parameter and the possibility of negative spread (multiple NCIs occurring without a negative marker) is summarized in Table ., replicated from Giannakidou (a) and Giannakidou and Zeijlstra (). Strict NCI is remarkably uniform, as we can see, and it includes Romanian negative concord, as well as Creoles such as Martinique Creole (Déprez and Martineau ). Giannakidou notes in connection to these data:

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi



 

Table .. The distribution of NCIs with the strict non-strict parameter in a number of languages discussed in the literature Language

Strict: PreV NCI þ negation

PostV NCI þ negation Pre-V and Post-V nword

. Greek

Yes

Yes

No

. Hungarian

Yes

Yes

No

. Polish

Yes

Yes

No

. Russian

Yes

Yes

No

. Czech

Yes

Yes

No

. Serbian/Croatian

Yes

Yes

No

. Romanian

Yes

Yes

No

. Martinique Creole

Yes

Yes

No

 Catalan

Yes

Yes

Yes

. Spanish

No

Yes

Yes

. Italian

No

Yes

Yes

. Portuguese

No

Yes

Yes

. French

No

No

Yes

We have a continuum with Strict NC on the one end (rows –[]) and negative-spread-only on the other, identified with French in row [], which systematically licenses double negative readings. In between we have Catalan, closer to the Strict NC end, and Italian, Spanish, and Portuguese, closer to the French end. Keep in mind that the proper set of comparison here is the set of Germanic negative quantifiers which do not allow NC, and therefore answer ‘no’ to the three possibilities indicated in the table. French is almost identical to the Germanic situation, save for those cases where multiple occurrences of NCIs do not allow for double negative readings. (Giannakidou a: )

For the Romance family, we must say that (a) the sensitivity of NCIs to the presence of negation or another NCI looks gradient, and (b) if an NCI necessitates the exclusion of sentence negation (as Italian, Spanish, and Portuguese do) the co-occurrence with negation will trigger DN. Here are three generalizations that can be made about Romance NCIs: ()

Three types of Romance NCIs i. Negative NCIs (French) ii. Ambiguous NCIs (Spanish, Italian, Catalan, Portuguese) iii. Non-negative, NPI like NCIs (Romanian, Martinique Creole)

To make things a bit more complicated, notice that, as discussed in Giannakidou (a) and Déprez (), there are more subtle details to be addressed in Romance having to do

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi

     



with the internal composition of NCIs. For instance, although negative spread is fine with NCIs as independent arguments, it seems degraded when NCIs are used as modifiers or determiners. This fact has been noted in Acquaviva () for Italian (and attributed to Manzotti and Rigamonti ); there is more recent discussion in the literature (Déprez , Déprez and Martineau ): ()

??/*

Nessuno student ha letto nessun libro. nstudent has read nbook Int.: ‘No student has read any book.’

(Acquaviva : )

The sentence is reported impossible on the intended NC reading. It is only accepted with DN. Similarly, structures with two NCIs functioning both as determiners are excluded in Spanish and Catalan. Compare the two sets of sentences below: ()

a. Ningún estudiante dijo nada. N-student said n-thing ‘No student said anything.’

Spanish

b. Cap estudiant (no) va dir res. N-student not  say n-thing ‘No student will say anything.’ ()

a. # Ningún estudiante dijo ninguna pregunta. answered n-question. N-student

Spanish

b. Cap estudiant va contestar cap pregunta. N-student  answer n-question The reported unacceptable judgment in the Spanish (a) reflects the availability of only a DN reading. In Catalan, on the other hand, if we add the negative marker ‘no’, the result is an NC reading ‘No student answered any question’. Naturally, theories of NC should, depending on the language(s) of departure, be able to explain the differences in licensing conditions that NCIs are subject to cross-linguistically, and how the licensing correlates with intended negativity of NCIs. In strict NC languages it is easier to make the argument that NCIs are not negative, and we will start with these.

.. NCI  

.................................................................................................................................. In strict NC the NCI is an NPI and can never be used with negative meaning—apart from the context of ellipsis in fragment answers. But if fragment answers involve ellipsis (as has been accepted since Giannakidou , , a; see also Merchant ) then negation can be assumed to be present in the logical form. If NCIs are non-negative, absence of DN in strict NC languages follows without stipulations. NCIs as NPIs would still differ syntactically from any, as suggested by Zeijlstra (a, b) because NCIs, but not any, are required to stand in a syntactic Agree relation with negation.

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi



 

Ladusaw () was among the first to hypothesize that NCIs are non-negative NPIs, and Acquaviva (), Déprez (), and Giannakidou (, ) developed this idea further by arguing that NCIs are indefinites, that is variables without quantificational force. Giannakidou (, a) proposes another version of the non-negative analysis by arguing that some NCIs are universal quantifier NPIs. I will consider the indefinites proposal in this section.

.. The indefinites analysis: NCIs and any Ladusaw (, ) hypothesizes that we can reduce negative concord to an independently motivated mechanism: binding of indefinites. Acquaviva (, ), Giannakidou and Quer (, ), Giannakidou (), Déprez (, ), Richter and Sailer (), and others at the time implement the program in various ways, all of which share that NCIs are indefinites with no quantificational force of their own (Kamp ; Heim ): ()

[[n-person]] = person (x)

There is no negation in the meaning; NCIs are indefinites like a person, any person. And just like any, NCIs differ from regular indefinites in that they come with an NPI requirement (Ladusaw  calls it “roofing”) that they be bound by existential closure under negation. Even for Germanic negative quantifiers, analyses have been formulated as containing negation and indefinite (Rullmann a; Penka ; Merchant ): no one, k-ein can be understood as not one, not ein. Such analyses draw often on the presence of an indefinite (ein), and split scope readings (Penka ; Zeijlstra this volume); some ellipsis supporting facts are discussed in Merchant (). Merchant () is the most recent incarnation of the indefinites analysis of Afrikaans NCIs, arguing additionally for a difference in the type of the variable, and characterizing the NCIs as negative isotopes. The indefinite analysis is consistent with the fact that many NCIs morphologically reflect an existential component ‘one’: for example uno in Italian nessuno or enas in Greek kan-enas. NCIs can also have existential non-negative use in non-negative polarity contexts: ()

a. Li diràs res? Him/her tell.. n-thing ‘Will you tell him/her anything?’

Catalan

b. Si aneu enlloc, digueu-m’ho. If go. anywhere, tell.. me ‘If you go anywhere, tell me.’ c. Tothom qui vulgui res, que m’ho digui. Everybody who wants n-thing, that me tell ‘Everybody who wants something/anything, should let me know.’ ()

Todos aquel que tenga nada que dicer All who that have n-thing that say ‘Everybody who has anything to say . . . ’

Spanish

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi

      ()

a. E’ venuto nessuno? Is. come n-body ‘Has anybody come?’

 Italian

b. E l’idea più stupida che nessuno abbia mai avuto! Be. the idea more stupid that n-body have.. ever had ‘It’s the dumbest idea anybody ever had!’ Here we see NCIs in three languages appearing in non-negative non-veridical environments (questions, conditionals, scope of universals) with a purely existential reading, similar to any. Greek has been argued to distinguish the indefinite NCI intonationally from the socalled ‘emphatic’ NCI (Veloudis ; Tsimpli and Roussou ; Giannakidou ). Chazikonstantinou  presents experimental evidence that the prosodic differentiation is real involving both lengthening and higher pitch. Here is the paradigm more fully: ()

kanenas/KANENAS tipota/TIPOTA pote/POTE puthena/PUTHENA katholu/KATHOLU

‘anyone, anybody/no-one, nobody’ ‘anything/no thing’ ‘ever/never’ ‘anywhere/nowhere’ ‘at all/not at all’

One element in the paradigm contains the word kan, which is one of the four words for ‘even’ that Modern Greek has (Giannakidou ; Giannakidou and Yoon ). Kan-enas is literally ‘even one’, but notice that this composition does not characterize the entire paradigm (see Xherija , Giannakidou and Yoon  for Korean and Albanian). One of the items contains the word ‘all’, as we can see. With negation and antiveridical without, both variants of NCIs appear: ()

a. Dhen idhe kanenan o Janis.  saw n-person the John ‘John didn’t see anybody.’ = John DIDN’T see anybody (#at all). b. Dhen idhe KANENAN o Janis.  saw n-person the John ‘John didn’t see ANYBODY at all.’ c. *Idhe kanenan/KANENAN o Janis. saw NPI-person /n-person the John

()

xoris na dhi {kanenan/KANENAN}. without  see. n-person ‘without having seen anybody.’

Antiveridicality is a property of negative expressions (Giannakidou , , ), and ‘without’ is an antiveridical licenser of NCIs generally, while it is morphologically not negative. Without cognates do license NCIs in the other languages discussed so far; it thus becomes apparent that negative concord is antiveridical concord (see Giannakidou ,

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi



 

, for more specific discussion), not simply negative. In many languages, NCIs do appear after words meaning BEFORE as long as they receive antiveridical reading (as in John died before he saw any of his grandchildren). Truth-conditionally, the statements with emphatic and non-emphatic NCIs are equivalent, but the emphatic variant is scalar (see Giannakidou and Yoon ), whereas the nonemphatic is not. Importantly for our discussion, the two paradigms also differ with respect to apparent negativity and distribution, as discussed in great detail in Giannakidou (, , a). In non-negative contexts, the emphatic NCI is ruled out: ()

Pijes {pote/*POTE} sto Parisi? went. n-ever in-the Paris ‘Have you ever been to Paris?’

Hence Greek distinguishes the any-like NCI (though the non-emphatic NCI lacks the free choice reading of any) from the emphatic NCI which is the form used in the fragment answer: ()

Q: Pjon idjes? ‘Who did you see?’ A: {KANENAN/*kanenan} ‘Nobody/*Anybody’.

Giannakidou (, a) argues that the NCI in fragment answers is the remnant of an elliptical structure, and “given that the remnants in fragment answers are accented, nonemphatics are excluded because they are not accented” (Giannakidou a: ). The ability of an NCI to give a negative fragment answer is therefore not a criterion for negativity, since the structure contains ellipsis that itself contains negation. For more on the issue of ellipsis and NCIs see Giannakidou (, a), Watanabe (), and Zeijlstra () with some critical remarks. Merchant () offers further supporting comments on the ellipsis account of NCIs. The Greek data suggest that under negation there are multiple strategies: one with an NCI (emphatic), and one with an NPI (non-emphatic kanenas). Many Slavic languages also have existential NPIs and NCIs, but unlike Greek, they block the existential NCI in the context of negation. Consider the distinction between i-NPIs and ni-NPIs in Serbian/ Croatian discussed in Progovac () and Blaszczack (); the phenomenon is known as the bagel problem, after Pereltsvaig (). It appears that the NPI strategy is also blocked in Romance languages (though it is present in French based Creoles, as we know from Déprez’s data). Greek, finally, also allows us to see that prosody plays a role in distinguishing strategies with negation.² Crucially, the recent works on double negation (Déprez et al. ; Espinal et al. ; Espinal and Tubau a) also point out a role of prosody in

² Prosodic differentiation is employed to distinguish NPI paradigms more generally (see Hoeksema  and Krifka’s  emphatic and non-emphatic any). We find similar observations for Arabic NPIs (Hoyt ).

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi

     



distinguishing DN from concord readings (see also Espinal and Prieto ; and Espinal and Prieto, this volume). The indefinites approach also faces challenges, noted early on (Déprez ; Giannakidou and Quer ; Giannakidou , , a). Unlike indefinites which have relatively free scope, NCIs are generally not licensed long-distance (see also Zanuttini ; Longobardi ; Progovac ; Przepiorkowski and Kupc ; Brown ; among many others). The Greek examples below allow us to see the difference: ()

Dhen prodhosa mistika [pu eksethesan {kanenan/* KANENAN}]  betrayed. secrets that exposed. n-person ‘I didn’t reveal secrets that exposed anybody.’

Greek

The non-emphatic indefinite behaves like any, but the emphatic NCI is blocked. Likewise in islands: ()

I Ariadne dhen ípe oti ídhe {tipota/*TIPOTA}. the Ariadne  said. that saw. n-thing ‘Ariadne didn’t say that she saw anything.’

()

a. Milan ne tvrdi [da Marija poznaje {*ni(t)koga/i(t)koga}] Milan  claim that Maria know n-person ‘Milan does not claim that Maria knows anybody.’

Serbian

b. Ne zelim da vidim {*ni(t)koga/i(t)koga}  wish. that see. n-person ‘I do not wish to see anybody’ In the Serbian examples (from Progovac ), the so-called i-indefinites are just like any in being licensed long distance. The NCI ni-indefinite, on the other hand, is like the Greek NCI, licensed only locally. In Polish and Russian locality is very strict: NC is impossible in all non-monoclausal domains, even if these domains are subjunctive-like or infinitival (where Greek and Serbian NCIs would be allowed). Other languages allow long-distance licensing of NCIs in subjunctive clauses (e.g. Spanish and Italian, cf. Herburger , Zeijlstra a; Greek, Giannakidou , a). If NCIs were simply indefinites, we should not find such locality constraints, as indefinites have more free scope, as observed above with any/ non-emphatics. Syntactically, the difference can be understood as coming from two different mechanisms: NPI licensing in the case of the any-like NCI, but the emphatic variant might be further constrained for some other reason, for example quantifier movement (as Giannakidou , a argues, to be discussed soon).³

³ Giannakidou and Yoon () suggest tentatively that intonation could be seen as the realization of a σ feature in the sense of Chierchia (), and this feature requires local licensing. If that is the case, it must be admitted that any lacks the σ feature, pace Chierchia, since any is licensed non-locally; see Giannakidou () for a more comprehensive criticism of the σ-approach.

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi



 

... Indefinites plus agreement Zeijlstra (a, b) presents a variant of the indefinite approach, partly motivated by the locality considerations mentioned above. He proposes that NCIs are polarity indefinites carrying a syntactic feature [uNEG]. A similar idea has been proposed for NPI-even in Giannakidou (). The feature [uNeg] must be checked against a higher element carrying an interpretable negative feature [iNEG]. This view can be seen as an update of the earlier agreement approaches of the nineties (Zanuttini ; Haegeman and Zanuttini , ; Ladusaw ; Haegeman ); but unlike these approaches, for Zeijlstra and Giannakidou the NCI and NPI-even are not semantically negative ([uNeg]), hence there is no absorption.⁴ Zeijlstra further assumes, drawing on a parallel with pro drop, that as in other cases of agreement, the element carrying the meaningful, that is interpretable feature may be phonologically null. As in pro drop, the NCI and the finite verb have a feature requiring that some other element check it (the abstract negative operator and pro, respectively), as suggested below (from Giannakidou and Zeijlstra ): ()

a. OpNEG[iNEG] nessuno[uNEG] telefona b. pro[i3SG] telefona[u3SG]

Italian

If the negative marker carries a feature [iNEG], no NCI is allowed to c-command it, as feature checking cannot take place in the reverse direction. In strict NC languages, the negative marker is preceded by an NCI. For Zeijlstra, this indicates that in these languages the negative marker cannot carry [iNEG], which means that it cannot be the vehicle of semantic negation. Sentential negation in strict concord is therefore an NCI with [uNEG] itself, and must check this feature against a phonological null negative operator. Below are the relevant structures for preverbal and postverbal NCIs, for strict and non-strict NC: ()

a. OpNEG[iNEG] nikdo[uNEG] ne volá[uNEG] b. OpNEG[iNEG] ne volá[uNEG] nikdo[uNEG]

()

a. OpNEG[iNEG] nessuno[uNEG] telefona b. Non[iNEG] telefona nessuno[uNEG]

()

Dnes. Op¬[iNEG] nevolá[uNEG] today .calls ‘Today he doesn’t call.’

Czech, Strict NC

Italian, Non-strict NC

Czech

Negative markers in non-strict NC languages (Italian non) carry a feature [iNEG] whereas negative markers in strict NC languages carry [uNEG]. This says essentially that some negative markers are meaningful (those with [iNEG]), while others are meaningless

⁴ Zeijlstra’s system allows for Multiple Agree (cf. Ura ; Haraiwa ), and multiple n-words can be checked against a single negative operator; Haegeman and Lohndal () have an alternative without Multiple Agree.

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi

     



(those with [uNEG]).⁵ The distinction, however, seems entirely arbitrary; and without the NCI, we are forced to assume that the actual semantic [iNEG] negator is phonologically null. This is in conflict with what we know about negation, namely that it is always marked (Horn a), and also raises questions about what children learn when they learn to use negative makers, which is quite early. One potential advantage of Zeijlstra’s system is that it alludes to an independently motivated self-licensing principle (whose application is attested in other Agree domains as well), and explains the locality of NC, which is a challenge for the indefinites approaches. Zeijlstra’s approach allows dissociation of the semantic content from the morpho-syntactic make-up of NCIs, a stance that can be traced back to Giannakidou’s idea that PIs carry sensitivity features: I will assume that polarity items are ‘special’ expressions in that they encode a sensitivity feature. Sensitivity features are semantic features, part of the lexical representation of the polarity items, and they encode the semantic ‘deficiency’ of these items. Sensitivity features are present in the lexical semantics of the PIs at least at an abstract level, and they may, but need not, correspond to syntactically active features. In some cases, they are indeed realized as morphological features, in other cases, as phonological features (for instance with emphatic NPIs in Greek). (Giannakidou : )

In the NCI, Zeijlstra argues, the sensitivity feature is syntactically realized as [uNEG]. EVEN in NPIs and NCIs may be another realization of [uNEG], as for example in East Asian languages (Korean -to NCIs), NPI-even (Giannakidou ), and Albanian NPIs (Xherija ). In languages without NC, Neg features are not part of the set of active syntactic features, and negative dependencies are thus not encoded syntactically. Collins and Postal’s (, ) idea that NPIs have a NEG feature is also in the same spirit, as well as Chierchia’s () treatment of NCIs (which is in spirit similar to Zeijlstra’s); for acquisition data supporting [Uneg], see Lin ().

.. NCI   

.................................................................................................................................. Giannakidou (, a) suggests that one of the strategies for negative concord must include a universal quantifier above negation: ()

Logical representations of general negative statements (i) 8x[P(x) ! ¬ Q(x)] (Universal negation) (ii) ¬9x[P(x) & Q(x)] (Existential negation)

⁵ Zeijlstra (b) argues that French pas is an example of a negative marker that lacks the Neg feature. It is only semantically but not syntactically negative. As a consequence, French pas cannot participate in concord relations. This obviously cannot be true of Quebecois or French based Creoles where in fact pas does participate in negative concord (Déprez ; Déprez and Martineau ).

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi



 

In languages without NC like English, these two scopes manifest as anybody and nobody; in an NC language such as Greek, if the non-emphatic reflects the existential construal, as we illustrated earlier, the emphatic NCI can be thought to reflect the universal construal. Suranyi () reports likewise two paradigms of Hungarian NCIs parallel to Greek (see also Szabolcsi ). In an earlier paper by Szabolcsi (), Hungarian NCIs were uniformly analyzed as universal quantifiers. Certain languages, such as Hebrew, employ morphologically universal NCIs and we find similar negation with universals also in Ancient Greek (Giannakidou a; Chatzopoulou ). Universal NCIs are found also in East Asian languages (for universal analyses of Japanese and Korean NCIs, see e.g. Shimoyama ; Sells ; Yoon ). Giannakidou (, a, a) motivates the universal approach with a number of tests, including scope parallelism between NCIs and universal quantifiers (relying on earlier work by Farkas and Giannakidou ), and the locality of NC. Giannakidou summarizes the following criteria for universal NCIs: ()

Diagnostics for a universal NCI A universal NCI has the following properties: (a) It is licensed only by local negation; long-distance licensing may be allowed only through a transparent domain (infinitival or subjunctive clause). (b) It can be modified by modifiers corresponding to almost/absolutely. (c) It cannot bind donkey pronouns. (d) It cannot be used as predicate nominal. (e) It expresses existential commitment, that is we tend to interpret it with a nonempty restriction, like with other universals. (f) It can be used as topic in topicalization structures, that is coindexed with a clitic pronoun (or a pronoun performing the respective function, if a language does not have clitics). (Giannakidou a: )

Giannakidou and Zeijlstra () reproduce part of that discussion, so I will repeat the arguments here. The universal analysis says that the NCI is an NPI that, for licensing, must outscope negation to achieve the right interpretation. This means that at the level of LF, the universal NCI must precede the negative marker; postverbal NCIs only do so after spell-out. Licensing, therefore, in this case is anti-scope: the licensee (NCI) must take negation in its scope. This account is compositional, as the only negation comes from the sentence negator. NC universals are different from negative universals such as nobody in that the former can only combine with negative predicates. Szabolcsi (: –) implements a similar derivation for Hungarian NCIs, and Merchant () encodes an analysis similar in spirit for Afrikaans NCIs as negative isotopes. The LF derivation for ‘KANENAS dhen irthe’ Nobody came is as follows, from Giannakidou (a): ()

a. b. c. d.

[[ KANENAS]] = λP λQ. 8x person(x) ! Q(x) [[ dhen irthe]] = λy. (not came (y)) [[ KANENAS]] ([[ dhen irthe]]) = λP λQ. 8x person(x) ! not came (x) [KANENAS1 [NegP dhen [TP irthe t1]]]

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi

     



In such configurations, KANENAS undergoes QR past dhen and lands in Spec, NegP, through an orthodox implementation of QR as adjunction (May ), in this case to NegP (Giannakidou a: ). Multiple occurrences of emphatic NCIs require successive adjunctions to NegP. Recall that no double negation reading arises. In principle, the number of emphatic NCIs allowed is unlimited, as can be recalled: ()

Dhen ipe KANENAS TIPOTA.  said. n-person n-thing ‘Nobody said anything.’

NCI universals are NPIs because they need negation for licensing—recall that Greek, Japanese, and Korean are strict NC languages. But universal NCIs must also reach a position outside the scope of negation; therefore, licensing for universal NCIs should be understood as anti-scope syntactically. NCI universals pose selections on the type of the predicate they combine with, that is they can only combine with negative predicates, and this translates in more recent work by Merchant to the concept of negative isotopes.

.. NCI  

.................................................................................................................................. The first negative analyses are found in the Neg-criterion approaches of Haegeman and Zanuttini (, ) and Zanuttini (). These early approaches rely on a parallelism between wh-dependencies (the wh-criterion) and negative concord. The idea that NCIs are negative translates into, in some form or other, absorption. More recent developments such as Déprez (), de Swart and Sag (), and de Swart () talk about ‘resumption’. Arguments against equating absorption with resumption, and more generally against the view that negative concord strictly resembles wh-dependencies are provided in Déprez (). Déprez (, , ) proposes to use resumptive quantification to account for French (but not Haitian Creole) NC readings. For Déprez, it is the differences between the nature of the NCI that justify the fact that resumptive quantification is only used with certain types of NCI and not others. Déprez (, ) proposes that to allow resumptive Quantification, an NCI must have the same nature as a numeral quantifier like zero, that is it must have quantificational force and a numeral nature. While French NCIs are of this nature, this is not true either of English or German negative quantifiers or creole NCIs which are indefinites akin to bare NPs in nature with no quantificational force. This explains why only French seems to have a freer alternation between DN and NC readings. The zero analysis of NCIs is essentially a variant of the indefinites analysis, and it is difficult to distinguish empirically what it means to be a zero indefinite versus a negative quantifier—or, as mentioned earlier, a Germanic indefinite that consists in NEG- ein (k-ein, g-een, etc.) Negative absorption mechanisms, on the other hand, allow a series of n unary negative quantifiers to be interpreted as a single negative quantifier. For de Swart and Sag (), and de Swart (), NC is derived in a way similar to the pair-list readings of multiple wh-questions: ()

Who bought what?

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi



 

The most salient reading here is a reading where it is asked what multiple pairs there are, such that x bought y. An answer could be, for instance, John bought apples and Mary pears. This is a reading where a single wh-operator binds a pair of variables, similar to unselective binding; syntactically, then, the similarity is obvious with the indefinites approach. In the French example below we observe two readings: one in which every negative quantifier binds a single variable, and one in which a single quantifier binds a pair of variables. The former yields DN, the latter concord: ()

Personne (n’)a téléphoné personne. N-body has called n-body (i) DN: No one is such that they called no one ¬9x¬9y Call(x, y) ‘Nobody calls nobody.’ (ii) NC: No pair of people is such that one called the other ¬9 Call(x, y) ‘No one calls anyone.’

The mechanism responsible for NC is in effect indistinguishable from unselective binding. But by allowing also the purely negative construal, this type of theory systematically generates two readings: a double negative (DN) and an NC reading. This, however, is too strong a prediction. As said earlier, with strict NC, DN is either unattested (or marked with distinctly different intonation); and in Catalan NC is preferred, as recent experimental work has shown (Déprez et al. ; Espinal et al. ; Espinal and Tubau a). Therefore, replacing unselective binding with negative absorption is not justified for a significant portion of languages, including some Romance. There is an intuition that in concord languages the DN reading is marked, while the concord is the default. For de Swart and Sag (: ), this is “really a question about the relation between language system and language use.” In principle, both interpretations are always available, but which reading surfaces is a matter of ‘use’. I find this to be too weak an explanation, as very few languages are like French where DN and NC readings are allegedly free. NC seems to be the default reading for almost all NC languages, and for French more empirical work is needed to be sure that the two readings are equally available. In Germanic, where NCIs are unquestionally negative, NC never arises. This contrast is in need of explanation, and again suggests that the two readings are not symmetrical: ()

Jan belt niet niemand. Jan calls  -.body DN: ‘Jan doesn’t call nobody.’ = ‘Jan calls somebody.’ *NC: ‘Jan doesn’t call anybody.’

Dutch

It has to be stipulated that the Germanic negative quantifier does not undergo absorption— but this doesn’t seem to follow from any other difference between the two, as would be desirable. Finally, the role of negative markers and the absence of negative spread in strict NC patterns remain unclear if NCIs are negative. If NCIs are negative quantifiers, why can they

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi

     



not stand by themselves in strict NC? Strict NC is, in my view, a serious problem for the negative theories of NCIs, absorption, and resumption alike. Why are negative markers obligatory? Why are they needed? De Swart and Sag argue that negative markers should be conceived as zero-quantifiers in the sense that they lack an argument. But assuming that negation is a ‘zero quantifier’ is at odds with the fact that negation is a logical connective, not a quantifier. We can’t simply trivialize the difference. And even if we were to grant this assumption, it would only explain why negative markers may participate in NC; why they are necessary in strict NC would still be left unexplained. I want to conclude this section with some comments on the possibility of multiple lexical entries of NCIs. The systematic coexistence of negative concord and DN readings in a single language suggests, more straightforwardly, that the NCI is ambiguous. Déprez (), Martins (), Herburger (), and others have proposed ambiguity accounts. Regarding Catalan in particular, one can construe two varieties—a polarity indefinite NCI and a negative quantifier, as suggested below (Espinal and Tubau a: ()): ()

a. NCI1: polarity indefinite meaning characterized [þσ]. b. NCI2: indefinite negative quantifiers

In recent experimental work, Espinal and Tubau (a) and others mentioned earlier propose that there are two competing lexical variants of NCIs in Catalan and Spanish: a polarity variant and a negative existential quantifier variant (like kein).

.. C

.................................................................................................................................. I will conclude, echoing Giannakidou and Zeijlstra () that it is ‘no simple business’ to say whether n-words (NCIs) are or are not negative, and whether being negative translates into having NEG features. The Greek polarity indefinite kanenas, for instance, has no NEG feature. I hope to have shown that some generalizations are indeed clear, and allow clear predictions. First, it is most helpful to think of negative concord and polarity as related dependencies under negation, manifesting themselves as different strategies within and across languages. There is an NPI strategy (clearly illustrated with the Greek non-emphatic NPI and any), and this strategy does not involve NC. There is also the negative concord strategy with an NCI. In languages lacking NCIs, negation creates no dependency other than the NPI one. NCIs can be used to perform multiple strategies if necessary, as appears to be the case in Romance. In languages like Greek, the polarity strategy and the concord strategy are distinguished intonationally. In Slavic languages, there is no polarity strategy under negation, only the concord strategy. Thinking in terms of a family of strategies is, I believe, most productive, and helps put a lot of diverse phenomena into a manageable landscape. Second, in strict NC, NCIs behave like NPIs and are arguably not negative. It is very tenuous to extend any version of the negative analysis to strict concord, without having to

OUP CORRECTED PROOF – FINAL, 5/3/2020, SPi



 

make ad hoc stipulations. The indefinite or universal analysis of NCIs in strict concord explains the distribution more straightforwardly, that is with much fewer stipulations. Third, syntactically, negative concord appears to be not a redundancy but a dependency of the kind we observe in language (i.e. indefinite binding, agreement, absorption, resumption, quantifier movement). Therefore, negative concord is not extraordinary but rather a member of the set of phenomena that reveal that the grammar of natural language has its own specific design and is not reducible to logical syntax alone.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  ......................................................................................................................

   ......................................................................................................................

̈    

.. W     ?

..................................................................................................................................

... Single negation and double negation readings Chapters , , , , and  in this volume introduce the theoretical background on negation markers, negative indefinites, and the syntax-semantics interface, with references to Baker (), Payne (), Corblin (), Horn (, a), Herburger (), Fălăuş (), de Swart (, ), etc. Depending on the language and the syntactic configuration at hand, a sequence of multiple negative expressions in the sentence can give rise to a single negation (SN) reading (either ¬p or ¬99), and a double negation (DN) reading (¬¬p or ¬9¬9). Compare French (a) and English (b):¹ () a. Elle ne vient pas. she  comes  ‘She didn’t come.’ b. Carol didn’t dare not to come. Carol did  dare  to come

written French SN DN

In the remainder of this chapter, we talk about sequences of negative expressions when dealing with multiple negative forms and DN when dealing with a double negation meaning.

¹ A note on glosses: =sentential negation, -body, -thing, etc. are negative indefinites in either double negation or negative concord languages.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



̈  

... Negative concord languages and double negation languages The frameworks introduced in Chapters , , , and  build on the differences between negative polarity items (NPIs) and negative concord items (NCIs) to establish a distinction between double negation and negative concord languages. Relevant references include Ladusaw (, , ), Zanuttini (), Vallduví (), Krifka (), Haspelmath (), van der Wouden (), Giannakidou (, a), Zeijlstra (a), Penka (), Collins and Postal (), and others. NPIs are non-negative expressions that typically occur under the scope of negation, for instance English anybody. NCIs like Italian nessuno are similar in form to the English negative indefinite nobody, but give rise to a single negation (SN) rather than a double negation (DN) reading when combined with negation or other negative indefinites. The examples in () (from de Swart ) illustrate: () a. Non ho visto nessuno.  has seen -body ‘I haven’t seen anybody.’

Italian SN

b. I haven’t seen anybody. ¬9x See(I, x)

English SN

c. Nobody said nothing. ¬9x¬9 Say(x,y)

English DN

Italian, French, and Spanish are Negative Concord (NC) languages, because sequences of negative indefinites give rise to a single negation (SN) reading (a). English, Dutch, and German are Double Negation (DN) languages, because they employ regular indefinites or NPIs under negation (b), and sequences of negative indefinites give rise to a DN reading (c). Note that the use of NPIs is obligatory in contexts like (b), because something is a Positive Polarity Item (PPI), an expression that refuses to take narrow scope under negation. Strict negative concord languages always require NCIs to co-occur with negation (standard French, at least in the higher registers). Italian and Spanish require the cooccurrence with negation only when the NCI appears in postverbal position, as in (a); this is called non-strict negative concord. For more discussion on the optional and obligatory use of NPIs and NCIs, see Labov (b), Haegeman and Zanuttini (), Szabolcsi (), de Swart (), and references therein. Although the contrasts are typologically strong, a closer investigation of negative concord languages reveals that a sequence of negative indefinites in an NC language sometimes allows a double negation reading. The observations go back to Zanuttini (), Haegeman and Zanuttini (), Corblin (), Depréz (), Herburger (), de Swart and Sag (), Borsley and Jones (), Fălăuş (), de Swart (), Puskás (). See (a) (from Corblin ) for an example. Similarly, a sequence of negative expressions can give rise to an SN reading in a DN language like Dutch, as in (b), from Zeijlstra (a).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  



() a. Personne n’est le fils de personne. -body  is the son of -body = No one is the son of anyone. = Everyone is the son of someone.

French

b. Zij heeft nergens geen zin in. she has -where -one lust in ‘She doesn’t feel like doing anything at all.’

Dutch

[NC] [DN]

[SN]

Corblin () emphasizes the role of pragmatic likelihood of the DN reading in (a), and Zeijlstra (a) argues that SN readings like the one in (b) are restricted to adjacent negative expressions and create special, emphatic negations. Both examples thus have a special status in their language.

... The role of prosody in the perception of double negation and negative concord Recent research by Blanchette () and Blanchette et al. () takes the debate one step further, and suggests that the DN/NC distinction is not operative at the macrolevel of languages, but is subject to microvariation across dialects and registers. Their focus is on English, a language that is known to have both DN and NC varieties. Their experiments show that speakers of Standard Average English (a DN variety) can detect SN readings in the presence of contextual and prosodic support. These results mirror the perception experiments carried out for Afrikaans by Huddlestone (). Standard Afrikaans assigns a DN reading to a sequence of two negative indefinites, but colloquial Afrikaans is an NC variety. Huddlestone found that speakers who produce standard Afrikaans can access the SN reading in a perception experiment. Further cross-linguistic evidence that prosody plays a disambiguating role comes from experimental research on French, Catalan, and Spanish as reported in Déprez et al. (), Déprez and Yeaton () and Chapter  in this volume. Unfortunately, the results of similar perception experiments carried out by Fonville () for Dutch were inconclusive. Furthermore, Blanchette’s results are duplicated for child English, but not adult English in Thornton et al. () (see also Chapter  in this volume). The role of acoustic cues in the detection of DN and SN readings may well be the main open question in current experimental research. The experimental literature raises the question when and where DN readings arise in either class of languages, and whether the grammatical distinction between DN and NC languages can be upheld. The aim of this chapter is to complement the focus on perception and prosody with a corpus approach that analyses patterns of language use. A parallel corpus allows us to compare the cross-linguistic production of negation within a single register.

... Patterns of language production: Monolingual and multilingual corpus research Fonville (), Fonville and de Swart () carried out monolingual corpus research on double negation configurations in the Corpus of Spoken Dutch to find out what

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



̈  

characterizes configurations like those in (b). The Corpus Gesproken Nederlands (Spoken Dutch Corpus, CGN) is an annotated corpus consisting of  hours of spontaneous speech (Boves and Oostdijk ). Within this corpus it is possible to perform a content search (e.g. to search for occurrences of a word), and to impose constraints on such a content search. These features make it possible to search for occurrences of two different negative indefinites near each other in the corpus. Searches were performed for all possible combinations of two negatives out of the following Dutch negative terms:² niemand (‘nobody’), niets/niks (‘nothing’), nooit (‘never’), nergens (‘nowhere’), niet (‘not’), geen (‘no’), within a one-word, two-word, and three-word distance from each other. With just over  instances, the most frequent combination is niet niks/niet niets ‘not nothing’. Combinations starting with niemand (‘nobody’) or nergens (‘nowhere’) show a very low overall frequency ( negative > zero (after Van Gelderen : )

The cline in (a) is involved in the development known as Jespersen’s Cycle (section .). In some languages, the source of the negative adverb is already an adverb. Cape Verdean Creole for instance has a sentential negator derived from such an adverb, ka < Portuguese nunca ‘never’, (). In this language, Jespersen’s Cycle has in fact progressed so far that ka has become a preverbal negator, taking the place that Portuguese não had. ()

el ka ta ba la he   go there ‘he doesn’t go there’

Cape Verdean Creole (Posner : ff.)

In some Niger-Congo languages, such as many Bantu languages, locative pronouns or adverbs and/or possessive adjectives have been described in the literature as sources of new negators (for an overview see Devos and van der Auwera ). Devos, Tshibanda, and van der Auwera () argue for one of the postverbal negative markers in Kanincin (Bantu LA), (a)p(a), (a), that it has developed from a locative affix (‘there’) of class , (b), via a minimizer use meaning roughly ‘a little’, (c).¹ ()

a. ki-naa-búl-ááŋ-áp mwáan Kanincin _-.-hit--_ .child ‘I have not hit the child’ (Devos, Tshibanda, and van der Auwera : ) b. náá-ték-àà-p .-put-- ‘I put there (on top)’

(Devos, Tshibanda, and van der Auwera : )

¹ Note the similarity with the minimizer/negator use of Portuguese lá ‘there’ in sei lá ‘I don’t know (at all), I have no idea’, lit. ‘I know there’. The other postverbal marker in Kanincin, derived from the class  marker ku- with a possessive agreement affix, was grammaticalized via a use as emphatic focalizer (‘at all’).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



  c. náá-búl-aaŋ-â-p .-hit---. ‘I have hit a little.’

(Devos, Tshibanda, and van der Auwera : )

The cline in (b), involving negative verbs, can take a number of forms, some of which will be addressed in section .. In Croft’s Cycle (Croft ; Kahrel ), a negative existential expression (NEGEX, roughly ‘there is not’) is recycled as a negative marker, like ikai (ke) in Tongan (a). But the source of a negative auxiliary (as in Finnish, ()) does not necessarily have to be originally an existential expression in the cline towards negative particles (Dahl ; Honti ). Besides, negative cleft constructions can grammaticalize into negative particles, as has happened with Sicilian neca, from ne é ca ‘it’s not that . . . ’ (lit. ‘ is that’), (b). ()

a. existential construction Na’e ‘ikai [ke] ‘alu ‘a Siale   go  Charlie ‘Charlie didn’t go’ b. grammaticalized cleft Neca ci vonsi jiri.  there wanted. go. ‘They did not want to go there.’

Tongan (Churchward : ) Sicilian (Cruschina ; )

The current chapter will first look at Jespersen’s Cycle in more detail, and then briefly turn to other cyclic renewals of the expression of negation.

.. J’ C

.................................................................................................................................. Jespersen’s Cycle describes the cyclic renewal of the expression of sentential negation by a new adverbial element, which typically grammaticalized from an originally argumental pragmatic reinforcer. It received its name from Dahl () after the original observation by Jespersen (): The history of negative expressions in various languages makes us witness the following curious fluctuation: the original negative adverb is first weakened, then found insufficient and therefore strengthened, generally through some additional word, and this in its turn may be felt as the negative proper and may then in course of time be subject to the same development as the original word. (Jespersen : )

Different authors have proposed different subdivisions of this development, with between three and five clearly definable stages (e.g. Roberts and Roussou : –; Rowlett : ; Van Kemenade : ; van der Auwera : ; van der Auwera b: –; Willis a: ; Jäger : ; Willis et al.  b: ). The schema in () illustrates the cyclical

    



formal reduction and reinforcement in five stages; three stages would correspond to the surface constructions (e.g. I—ne, II—ne . . . pas, III—pas), without taking into account the varying optionality or availability of more than one construction at a given stage.² () I – clitic

IIa – clitic (+ free morpheme)

IIIb – free morpheme

IIb – clitic + free morpheme

IIIa – (clitic +) free morpheme

Some languages, like the Romance languages, seem to have undergone several cycles (or rather, spirals, as observed by von der Gabelentz  []: —we will stick to the more common terminology here) of reinforcement and weakening of negative markers (e.g. Schwegler ; Detges and Waltereit ; Hansen : ). In Latin, the negative marker non arose through reinforcement by oenum > unum ‘one’ (a), a process which repeated itself in the history of French in the reinforcement of the expression of negation by the minimizer ‘step’ (b). ()

a. archaic Latin classical Latin

ne oenum dico ‘ one say.’ non dico ‘ say.’

b. Old French Classical French Modern French Colloquial French

jeo ne dis je ne dis pas je (ne) dis pas je dis pas

‘I  say’ ‘I  say step > ’ ‘I () say ’ ‘I say ’

While it is geographically widespread in north-western Europe, and has been considered an areal feature by typologists (Bernini and Ramat ), this development is also found in Greek (Kiparsky and Condoravdi ; Willmott ; Chatzopoulou ), Niger-Congo languages (Güldemann ; Beyer ; Devos and van der Auwera ; and references cited in Devos, Tshibanda, and van der Auwera ), Afro-Asiatic languages (Gardiner ; Chaker and Caubet ; Lucas ), and creole languages (e.g. Posner ; Hagemeijer ).

² More will need to be said about the underlying structural analysis of the elements involved, whether they are optional or not. We return to this in section .. below.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

... Common developments As outlined above, Jespersen’s Cycle is characterized by the overlap of several grammaticalization clines leading to the (repeated) demise and renewal of the (adverbial) expression of standard negation. While languages differ in points of detail, about which more below, there are a number of properties of Jespersen’s Cycle that all languages undergoing it share. In all cases, an original negation particle is first reinforced by a new expression, and ultimately replaced by it. Typically, the reinforcing expressions which start to compete with the original expression of sentential negation at the beginning of Jespersen’s Cycle have a scalar meaning when they enter the cycle (Horn ; Eckardt , ; Breitbarth, Lucas, and Willis ). New elements feeding into Jespersen’s Cycle are either minimizers like ‘a drop’, ‘a crumb’, or ‘a dot’, indicating that a state of affairs does not hold even to the least extent, or generalizers like ‘not (even) one’ or ‘not (a single) thing’, indicating that not a single member of a relevant set is affected by a state of affairs. As seen above in relation to (), other types of elements can become new negators under Jespersen’s Cycle, such as locative or possessive markers. Also in those cases, however, an intermediate minimizer stage is attested. While generalizers are by their nature compatible with a wide range of predicates, minimizers are typically originally lexically restricted to specific predicates, see (a) vs. (b) (also, see Breitbarth, Lucas, and Willis ). ()

a. not + rain/drink/spill/cry/ . . . + a drop b. *not + sleep/wait/call/ . . . + a drop

Due to their scalar nature, such elements can serve to (emphatically) reinforce the expression of sentential negation (Horn ; Eckardt , ). In an affirmative clause, an element expressing a low point on a pragmatic scale (e.g. a drop > a cup > a bucket > a barrel > . . . ) is not very informative. But due to scale reversal under negation, such elements become very informative, as they carry a conversational implicature (also called a scalar implicature) that the situation described did not only not take place at all, (). ()

I did not drink a drop = I did not drink so much as even the smallest possible amount +> I did not drink (at all)

This means that initially, negation reinforcers serve to lay emphasis on the negative force, although van der Auwera (, b) has argued that there are some languages in which a reinforcer of the expression of negation has arisen from a non-emphatic element. In the cases he describes, however, the reinforcer is typically a doubled negative marker, that is, elements that already expressed negation before being used as reinforcers. Arguably, most if not all languages have ways of pragmatically emphasizing the polarity of negation, typically by exploiting pragmatic scales. The question rises, then, of when one can speak of the beginning of Jespersen’s Cycle. Both semantic and syntactic changes are involved in the grammaticalization of new negative particles from scalar reinforcers. Minimizers first need to undergo semantic bleaching of their original lexical restriction.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

    



Syntactically, both minimizers and generalizers need to lose their original syntactic status. Reinforcers that started out as argument of the verb are often initially restricted to transitive verbs, (). ()

We didn’t see / *care a living soul.

In the next step, they become negative polarity adverbs, which means that they are usable with all verbs. This syntactic change is facilitated by the availability of bridging contexts, in particular verbs with ambiguous argument structure, such as ‘profit’, ‘succeed’, or ‘care’, which allow an optional argument expressing the extent or degree of the profit, success, or care, () (Breitbarth, Lucas, and Willis ). ()

Ny thalwys idaw hynny dim . . . (Middle Welsh)  pay.. to. this anything ‘And this didn’t help him (at all) . . . ’ (Ystorya de Carolo Magno ., from Breitbarth, Lucas, and Willis : )

Before becoming syntactically completely unrestricted negative polarity adverbs, they may go through a stage in which they can be used as optional extent arguments (a), but cannot co-occur with transitive verbs (b), where they would compete with the theme argument. ()

a. No-l lavora gnente. -he work.. nothing ‘He doesn’t work.’ b. *No-l leze gnente libri. -he read.. nothing books ‘He doesn’t read books.’

Venetian

(Poletto : –)

After overcoming any transitivity restrictions, the pragmatic reinforcer becomes a negative polarity adverb reinforcing the expression of sentential negation. This point, at which a universally available negative polarity adverb becomes available, is probably the earliest one at which one can classify a language as having entered Jespersen’s Cycle. At this stage, the new reinforcer may still be subject to pragmatic and information structural restrictions, as argued for a number of incipient negators in Romance languages at different historical stages (Cinque ; Schwenter ; Visconti ; Hansen ), and for historical English (Wallage ). Once all context restrictions are overcome, the negative polarity adverb may be considered fully grammaticalized as the new neutral/standard expression of sentential negation. At the same time, the original negation marker needs to lose its ability to express sentential negation on its own, as there can only be one logical negation in one clause. Besides such semantic bleaching, it typically also tends to undergo formal changes, losing its syntactic or prosodic independence while changing from a free morpheme to a clitic, as can be illustrated with the development from Old Dutch to present-day (Standard) Dutch in (). The free (preverbal) marker ne in Old Dutch (a) becomes a clitic in Middle Dutch (b), often reinforced by the free (postverbal) particle niet (c). In Modern Dutch, only the free particle niet remains (d).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

()

a. free particle (non timebo quid faciat mihi caro) ne sal ic forhtan uuad duo mi flēisc Old Dutch  shall I fear what do me flesh ‘I will not fear what flesh can do to me.’ (Wachtendonck Psalms LV.; Breitbarth a: ) b. clitic Ghi hout u spot. I-n (=ic+en) doe you hold your mockery I- do ‘You are mocking. I do not!’

Middle Dutch (Beheydt : )

c. combination of clitic and free adverb hie-ne sal niet delen an den loon Middle Dutch he- shall  share in the reward ‘he shall not receive a share of the reward’ (Corpus Gysseling; C, Brugge, //) d. free particle Ik weet het niet. Modern Standard Dutch I know it  ‘I don’t know.’ After the grammaticalization of not, English has gone on to develop a more asymmetric negation strategy as a result of the establishment of auxiliaries as the only verbs able to move across sentential adverbs, including negation, and the concomitant rise of do-support in negative clauses: the form of a negative clause is not necessarily identical to the form of an affirmative clause minus a negative marker, as the finiteness is no longer expressed on the lexical verb in negative clauses, (b–c). ()

a. he knoweth / knows not Early Modern English b. he does not know / he shall not know / he has not known c. he doesn’t know / he shan’t know / he hasn’t known

Despite the common developments all languages undergoing Jespersen’s Cycle share— grammaticalization of a new expression of standard sentential negation out of a reinforcer with certain pragmatic properties, loss of the original negative marker, possibly repetition of these processes—, languages differ for example regarding the duration of the transition between the stages, and regarding the fine structure of the middle stage (which of the two elements is the main/standard expression of negation), and also regarding the fate of the original marker after establishment of the new one (see also section ..).

... The direction and speed of the change As alluded to in section .., all languages have ways of emphasizing the expression of sentential negation by means of scalar nominal elements, without these necessarily going on to become new negative markers. What are the triggers for an emphasizer to overcome initial semantic or syntactic restrictions, and what are the triggers for an original negator to

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

    



cease to express sentential negation? Essentially, two kinds of views regarding the direction of Jespersen’s Cycle can be distinguished in the literature: ones that see the development as a pull-chain, like Jespersen’s original analysis, and ones that see it as a push-chain. While Jespersen’s often-cited quote (see section .) suggests that the old marker first needs to ‘weaken’ before a reinforcer is put in place, hence favoring a pull-chain-scenario, the question of which development comes first—the weakening or the reinforcement—is far from settled. Many languages seem to be in no hurry to establish a systematic reinforcer despite phonologically and syntactically very ‘weak’ expressions of standard negation (see also Posner  for Romance varieties), such as the bound negative morpheme ne- in Czech, (). ()

Milan ne-volá Milan -call ‘Milan doesn’t call’

Czech (Zeijlstra a: )

It is therefore crucial to determine what ‘weakening’ means; van der Auwera () for instance explicitly separates formal from semantic weakening (and strengthening). Formally, a negative marker can be analyzed as weakening when it changes from a free morpheme to a clitic, thereby losing its syntactic or prosodic independence, or when it is lost entirely. Semantically and pragmatically, the weakening may manifest itself in the loss of the ability to express sentential negation on its own. In terms of formal morpho-syntactic features, ‘weakening’ of a negative marker may mean it changing from an interpretable to an uninterpretable negation feature, which then requires licensing by a carrier of an interpretable negation feature (Zeijlstra a; Van Gelderen , ; Willis a). We return to this point in section .. below. Under a push-chain scenario, a new expression of sentential negation comes to compete with the original expression, and ultimately pressures it out of use. From a functionalist perspective, this can happen when inflationary use of a pragmatic emphasizer of the kind all languages have (see above) leads to a loss of the emphatic import, creating a new pragmatically neutral negator which makes the old one functionally redundant (Schwegler : –; Detges and Waltereit ; Kiparsky and Condoravdi ). From a formalist point of view, one could say that once a pragmatic reinforcer still subject to semantic or syntactic restrictions (see section ..) has been reanalyzed as a universally employable negative polarity adverb, further reanalysis as a neutral sentential negator becomes possible (e.g. Breitbarth, Lucas, and Willis ). There are also ‘hybrid’ scenarios, according to which the (morpho-syntactic) ‘weakening’ of the original marker and the establishment of a new negator, pushing the original element out of use, go hand in hand (Wallage , ; Breitbarth ; Willis : –). In any case, Jespersen’s Cycle progresses by a new element taking over the function of expressing standard sentential negation. Given that each negative marker evolves along its own individual grammaticalization cline, the fate of the original negator seems to be predestined to be lost—“zero” in Van Gelderen’s (: ) overview reproduced in (a) above, once a language enters the cycle. This is indeed what seems to have happened in the North Germanic languages, most West Germanic languages, and it is currently happening in those Romance languages undergoing Jespersen’s Cycle (which are in different stages of

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

progress). However, comparing the diachronies of different languages having renewed their expression of negation in this way, significant differences can be noted. While the transition from stage II to stage III may be completed within a generation, it may also take several centuries. The former scenario is for instance found in Middle English (Wallage , ) and in Middle High German (Jäger ); the latter for instance in French (Catalani ) and southern Dutch (dialects) (Beheydt ), both of which only started losing the original preverbal particle since the nineteenth century (for French, cf. Martineau and Mougeon ), after having coexisted with the new standard expression of negation (pas and niet, respectively) since the Middle Ages. This raises the question of what determines the speed of the transition. It also poses questions for theories assuming that the new negator functionally competed with and replaced the old one. If real diachronic stability exists, what causes it? There is only very little literature as yet about these questions. For some languages, such as Swiss French (Fonseca-Greber ), (), or varieties of spoken Belgian Dutch (Breitbarth and Haegeman ), (), where the use of the original negative marker, n’ in () and en in () settles in at a low but stable frequency, a form of exaptation has been proposed: by taking on a new function (roughly, polarity emphasis), the particle no longer competes with the standard expression of sentential negation. ()

S: et ben les répondeurs ça sert à quelque chose . . . non . . . ( . . . ) and well the answer.machines that serves to some thing . . . no . . . S: mais nous on a même pas de répondeur . . . mais papa but we we have really  of answer.machine . . . but dad il n’ en veut pas . . . he  of.it wants  (Fonseca-Greber : )

()

Ge zou lyk peinzen dat da Valère is. Mo t=en is Valère niet. you would like think that that Valère is but it= is Valère not ‘One would think it was Valère. But it wasn’t Valère.’ (From Breitbarth and Haegeman : )

Further research into more languages will afford insight into the determinants of the speed of change and diachronic and stability.

... Formal analysis Many theoretical approaches to the analysis of Jespersen’s Cycle have followed Pollock’s () original proposal (a) of assuming that negation is structurally encoded in a functional projection NegP. ()

a. [TP [NegP [AgrP [VP ]]] b. [NegP SpecNegP [Neg’ Neg⁰ [VP ]]]

The availability of a head and a specifier position (b) that comes with such an analysis strokes well with the grammaticalization cline of negative markers under Jespersen’s Cycle: negation particles that are free morphemes may be analyzed as occupying the specifier of

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

    



NegP, particles that are clitics or affixes can be analyzed as the head of NegP. Such Specifier-to-head reanalyses are rather common, cross-linguistically; complementizers, for instance, often derive from former relative pronouns (e.g. German dass or French que). Besides, the availability of two positions (the head and the specifier position) creates the space needed for the presence of two elements at the intermediate stage(s) of the cycle. Such accounts of Jespersen’s Cycle are proposed for instance by Rowlett (), Roberts and Roussou (), Roberts (), van Gelderen (, ), or Willis (a).³ As alluded to above regarding the number of different stages that Jespersen’s Cycle has been divided into in the literature (), more needs to be said about the underlying structural analysis. What superficially looks like stage II (old negator + verb + new element) may involve an argumental reinforcer (a), an adverbial emphasizer (b), or a sentential negator (c). ()

a. [TP Je [T’ [T nei=vaisj ] [NegP [Neg’ [Neg ti ] [VP tj [DP (un) pas ]]]]]] b. [TP Je [T’ [T nei=vaisj ] [NegP [Neg’ [Neg ti ] [VP [AdvP pas ] [VP tj ]]]]]] c. [TP Je [T’ [T nei=vaisj ] [NegP pas [Neg’ [Neg ti ] [VP tj ]]]]]

These elements are in structurally very different positions, as () illustrates. The superficial similarity is of course what makes the reanalysis possible. Regardless of superficial word order, one can only speak of stage II of Jespersen’s Cycle in a technical sense once the new element has become an adverb (as in (b)) universally available in negative clauses, without syntactic or semantic restrictions, that is, at the negative polarity adverb-stage (Breitbarth, Lucas, and Willis ). Only languages in which one reinforcer reached this stage seem to have proceeded to the following stages of JC, viz. the increasing optionality of the original marker and its ultimate demise. Under the most common type of NegP-accounts of Jespersen’s Cycle, the original negator starts out as the head of NegP at the beginning of the cycle (cf. also (a)), bearing an interpretable/valued negation/polarity feature (details depending on the concrete proposal). This reflects its formal ‘weakness’ (if already a syntactic head) at the same time as its semantic independence (it can still act as the standard expression of sentential negation). The new element bears an uninterpretable negation feature, and needs to be licensed by the negative head under Agree (Chomsky , ), (a). After reanalysis, the licensing relationship is reversed, (b).

³ There are some formal analyses of the syntax of sentential negation arguing, based on synchronic evidence, that there are several NegPs per clause, depending on e.g. the diachronic source or the concrete semantic import of different negation markers in a language, particularly in cartographic (Zanuttini ; Poletto , ) or nano-syntactic approaches (De Clercq , b). Besides, there are proposals assuming that the availability of NegP in a language depends on the stage of Jespersen’s Cycle the language is in (Zeijlstra a; Wallage ). Finally, it has been argued that a NegP is not necessary to account for Jespersen’s Cycle, and that differences in morpho-syntactic features and internal structure of negation particles are sufficient to account for the stages of the cycle (Breitbarth ). For the sake of simplicity, the most common analysis, using a single NegP, will be presented here, and the reader be referred to the literature cited for the alternative approaches.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 ()

  a. [NegP [AdvP pas u ] [Neg’ [Neg ne i ]]] ! b. [NegP [AdvP pas i ] [Neg’ [Neg ne u ]]]

This raises the following questions: I. How does the reinforcer acquire the uninterpretable negation feature requiring licensing by the negative head (original negative marker), and therefore getting it into a structural relationship with NegP? II. If there is an Agree relation between the old and new negative markers, whichever one of them licenses the other, why is the original negative marker lost once the new marker acquires the interpretable negation feature? Regarding (i), there are of course reinforcers in the history of several languages having undergone Jespersen’s Cycle that already express negation, viz. generalizers meaning ‘nothing’ or ‘never’ (cf. ()). Given that many languages at stage I of Jespersen’s Cycle seem to be negative concord languages, there is already an Agree relation between the reinforcer and the sentential negator.⁴ As for originally non-negative arguments such as French pas, they can be assumed to acquire the necessary [uNEG] feature by association with negative contexts. This can happen during language acquisition, where children may converge on a different grammar than the one(s) that generated their linguistic input (a.o. Kiparsky ; Andersen ; Lightfoot , ; Hale ). Such divergence is assumed to be driven by principles favoring more economical derivations with the same surface output (e.g. Roberts and Roussou ; Van Gelderen ). Such a principle, originally formulated in the context of language acquisition, not language change, is Schütze’s () Accord Maximization Principle, (). ()

Accord Maximization Principle Among a set of convergent derivations S that result from numerations that are identical except for uninterpretable phi- and case-features, such that the members of S satisfy other relevant constraints, those members of S where the greatest number of Accord relations are established block all other derivations in S. (Schütze : –)

() ensures that any relations that look like Agree, such as the customary co-occurrence of a negative marker and an adverbial reinforcer, will be interpreted as an instance of actual Agree, with suitable features.⁵ Regarding (ii), Van Gelderen’s (, , ) principle of Feature Economy, applied to the development of negation in (), introduces a trigger for the lexical renewal under Jespersen’s Cycle by postulating that uninterpretable features are more economical than

⁴ According to e.g. Rowlett () or Zeijlstra (a), all languages in which the negative marker is a syntactic head are negative concord languages. This correlation does not hold to % (see e.g. Breitbarth b), but may play a role in incipient Jespersen’s Cycle (see also van Gelderen : –). ⁵ This is similar to Hicks’s (: ) principle Maximize Featural Economy, “establish[ing] dependencies where possible.”

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

    



interpretable ones. The principle predicts that any element that has an [iNEG] feature will be prone to turn into one that has an [uNEG] feature, and eventually be lost. () Feature Economy Minimize the semantic and interpretable features in the derivation, e.g.: Adjunct/argument > Specifier (of NegP) > Head (of NegP) > affix semantic > [i] > [u] >— (van Gelderen : ) Furthermore, formal work on grammaticalization has correlated positions higher in the hierarchical clause structure with a higher degree of grammaticalization (cf. Roberts and Roussou’s  ‘upwards reanalysis’, or Van Gelderen’s  ‘Late Merge Principle’). Here as well, economy considerations are at play: movement is generally considered as ‘more costly’ than direct merger.

.. O  

.................................................................................................................................. The second type of developments leading to the renewal of the expression of negation mentioned in (b) above involves the recruitment of new negative markers generally speaking from a (negative) verbal expression. This type of development can take different concrete forms: (i) the grammaticalization of a new negator from the combination of an old expression of negation with an existential verb, the development of a new invariant negation marker out of a formerly inflecting negative auxiliary, (ii) the grammaticalization of negative markers out of prohibitive or negative verbs, and (iii) the grammaticalization out of former (negative) cleft structures. While these all at some stage involve a negated or negative verbal expression, they are different enough from each other not to be treated as instances of the same development, as there is not always a clear negative auxiliary stage. A development rather similar to Jespersen’s Cycle in the sense that a new standard negator replaces an old one after a period of emphatic use is what is known as Croft’s Cycle. The main differences between this development and Jespersen’s Cycle is that first, a negative existential expression arises, which is initially only used in existential clauses, before it spreads into non-existential negative contexts, eventually replacing the standard expression of negation there. Croft () observes that there are three attested types of interaction between verbal negators—the type of negators discussed in section .—and negative existential forms, (), and that in languages with synchronic variation, A varies with B, B with C, and C with A, suggesting a grammaticalization cline from Type A via Type B to C, and back to Type A, (). ()

Type A: verbal negator + existential predicate Type B: special negative existential predicate ≠ verbal negator Type C: special negative existential predicate = verbal negator



 

() Type A

Type B

Type C

For instance, the Syrian Arabic negator mā ‘not’, combines with verbs (a), also with the existential verb fī ‘there is’, (b) (Type A).⁶ ()

a. mā baʔref  I.know ‘I don’t know’

b. šu mā fī ħada bəl-bēt    someone home ‘Isn’t there anyone at home?’ (Cowell : –, via Croft : )

Croft (: ) cites Amharic as an example of a Type B language, (), where the standard verbal negator and the negative existential are formally distinct: ()

a. a-yalf-əm -pass...- ‘He doesn’t / won’t pass’

b. səkkwar yāllām sugar . ‘There is no sugar’ (Leslau : ; , via Croft : )

In Type C languages, the standard expression of negation is identical to the negative existential expression, as in Tongan () (Croft : ). ()

a. ‘Oku ‘ikai ha faiako !i heni.   a teacher at here ‘There isn’t a teacher here.’

b. Na’e ‘ikai [ke] ‘alu ‘a Siale  () go  Charlie ‘Charlie didn’t go’ (Churchward : –, via Croft : )

According to Croft, what leads from stage A to stage B is typically phonological fusion between the verbal negator and the existential expression. The change from stage B to stage C, that is, the change by which the specialized negative existential form becomes the expression of standard negation, seems to involve emphasis: at first, the negative existential form can be used as an emphasizer, as in Mara, (). ()

a. malu̥ y n̥a-ŋugu-ni gi-ŋara . ..-water- here ‘There is no water here.’

⁶ In those Arabic varieties that have also undergone Jespersen’s cycle, it is the bipartite negative marker mā . . . š (< mā . . . šayʔ ‘ . . . thing’) that combines with fī (Lucas , ).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

    



b. ganagu wu-nayi malu̥ y  /-saw . /  ‘He did not see him at all / He saw nobody’ (Heath : ; , via Croft : ) Besides the negative existential cycle, Croft also identifies a parallel negative prohibitive cycle, where instead of an existential verb, a (negative) imperative, irrealis or potential modal is grammaticalized as a new negative marker. The grammaticalization of negative markers from negative verbs such as fail (he failed to do that +> he did not do that) or die (Givón ) may also belong in this category (for an overview and references, see Van Gelderen ). A second development outside Jespersen’s Cycle is the negative auxiliary cycle (Dahl ; Miestamo ). Expressing negation by means of an auxiliary, itself probably originally created from a (negative) copula through Croft’s Cycle, is particularly characteristic for the Uralic languages (Honti ). As discussed briefly above in relation to (), negative auxiliaries take on person and number marking in negative clauses, while the lexical verb appears in a non-finite connegative form. Languages with negative auxiliaries vary with respect to the number of grammatical categories realized on the auxiliary. While Finnish for instance expresses person and number on the negative auxiliary, tense is expressed on the connegative verb, (). ()

a. Luen read..

b. En lue . read.

c. En lukenut . read..

Across the Uralic languages, there is a diachronic trend for tense, mood, and person/ number marking to be lost from the negative auxiliary. In some cases, they come to be expressed on the lexical verb. In the now-extinct Finnish dialect of Värmland (Sweden), the original negative auxiliary was reduced to a negative particle, while the former connegative was a fully finite verb () (Miestamo ). Other languages, such as Estonian, have not gone so far (yet); while the negative auxiliary has become invariant (Dahl : ; Honti : ), the lexical verb has not come to express person, number, or tense, (). () Ei minä lyö-n sinua  I hist-. you. ‘I will not hit you’ ()

a. Loen. read.. ‘I am reading.’

Ma ei loe. I  read. ‘I’m not reading.’

b. Loed. Sa ei loe. read.. you  read. ‘You are reading.’ ‘You aren’t reading.’

Värmland Finnish (Miestamo : ) Estonian

(Miestamo : )

While the juxtaposition of (), (), and () makes it look as though the negative auxiliary cycle proceeds from the grammaticalization of a fully finite negative auxiliary through several stages of transfer of finiteness categories to the lexical verb, this is not quite so deterministic. First, the Mordvin languages have for instance developed a complex variety

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

of negative markers, with negative paricles, nouns, and suffixes complementing the more traditional Uralic negative auxiliaries (Hamari ). Second, the case of Estonian shows that a language using the negative auxiliary strategy can also enter (incipient) Jespersen’s Cycle. The negative adverb mitte, which is normally the constituent negator in Estonian, is also used as an emphasizer in sentential negation (Ehala : –), (), as in early stage II of Jespersen’s Cycle (Tauli : ).⁷ ()

Ta ei lähe (mitte). he  go. . ‘He’s not going (at all).’

(Ehala : )

A final non-Jespersen’s Cycle type of renewal of the expression of standard negation is the grammaticalization of new negative markers from grammaticalized (negative) cleft structures. As with the negative existential and hortative cycles described by Croft () and the negative auxiliary cycle esp. in Uralic languages, this is a development that derives a new expression of standard negation from a higher negative verb. Cruschina () and Garzonio and Poletto (b) discuss one such development, the grammaticalization of Sicilian neca ‘not’ from a cleft structure (a), which in Standard Italian would be non é che ‘it is not that’. This element appears to have entered incipient Jespersen’s Cycle in the sense that it, like Italian mica, can now be used as a presuppositional negator, denying a contextual expectation to the contrary, (b). ()

a. Neca (< ne é ca) ci vonsi jiri. Mussomeli Sicilian  (<  is that) there wanted. go. ‘They did not wanted to go there.’ (Cruschina ; ) b. Sta lezioni neca si capisci. this lesson not  understands ‘(Contrary to what you think) one does not understand this lesson.’ (Cruschina : )

According to Garzonio and Poletto (b: –), neca has become an exponent of Focus in the clausal left periphery, which may be preceded by topics such as sta lezioni in (b), that is, the structure has become monoclausal.

.. C

.................................................................................................................................. Both Jespersen’s Cycle and Croft’s Cycle (in the more general sense of grammaticalizing higher negative verbs as standard negators) involve a replacement of an old negative

⁷ Estonian mitte derives from a partitive of ‘what’, that is, it is a kind of indefinite (Honti : ). As seen in section ., this is a common source of emphasizers in Jespersen’s Cycle; the West Germanic languages have for instance developed new negators from emphatic ‘nothing’. Its Finnish cognate mitään ‘nothing.PART’ is used as a presuppositional negation emphasizer (i.e. negating an expectation by an interlocutor), too (Willis et al. b: –).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

    



marker by a new one, via stage at which the new form has a special pragmatic function such as emphasis or presuppositional negation. As outlined above, several issues remain, such as the still ongoing debates about precise triggers of renewal, or about the reasons for differences in the speed of the change. The question of whether it is useful to unify the different types of development discussed in section . under a single term (e.g. negative auxiliary cycle) was left open given the great formal differences between for example the negative existential cycle, the renewal through the grammaticalization of negative verbs, and the renewal by reanalyzing former cleft structures as adverbs or particles negating finite verbs. The cyclic renewal of the expression of standard negation therefore continues as a fertile field of research.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ......................................................................................................................

                  ......................................................................................................................

 

.. I

.................................................................................................................................. T chapter is dedicated to two different, but often interconnected, diachronic processes that are known to affect negation systems: (i) the evolution of negative dependencies, that is the development of Negative Concord: this development consists in the establishment of a syntactic licensing relation between the negative operator and an indefinite determiner or pronoun in its scope. The indefinites entering Negative Concord (Negative Concord items) may have various historical sources (former negative polarity items, former negative indefinites, newly grammaticalized functional items). (ii) the grammaticalization of new items entering the negation system: systematic changes in the domain of negation are well known; examples come mainly from the diachronic evolution of negative markers (cf. Chapter  in this volume), but also narrow-scope existentials interacting with negation undergo systematic processes of formal renewal, substitution, and reanalysis. These processes will be reviewed in this chapter. Our understanding of the phenomena above has been substantially enhanced by the last few decades of diachronic research through the application of sophisticated theoretical models to historical data, and through a broader typological coverage (cf. Larrivée and Ingham ; Willis, Lucas, and Breitbarth b; Hansen and Visconti  for comprehensive overviews). This chapter summarizes the main insights of this research, and singles out some open issues and new questions. Section . is dedicated to the evolution of negative dependencies: after a short presentation of the categories assumed in the analysis (..), I discuss various aspects of the birth of Negative Concord systems: the interaction with morpho-syntactic processes affecting the negative marker (..); the role of the internal composition of pronominal items taking part in Negative Concord (..); the role of diachronic mechanisms of pragmatic strengthening involving the expression of focus (..). Section . deals with different

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



ways in which negative items emerge diachronically: I discuss the development of Negative Concord items from pronominal elements (..), their grammaticalization from nominal elements and the diachronic processes affecting minimizers and generalizers in the scope of negation (..). Section . summarizes the main conclusions and issues for further research.

.. T    

.................................................................................................................................. This section deals with the diachronic emergence of Negative Concord, the prime example of a syntactic dependency relation involving elements that express sentential negation. Negative Concord emerges once a number of syntactic prerequisites are in place: they will be reviewed in the next subsections. The indefinites participating in Negative Concord (Negative Concord items) are the result of various diachronic processes, which will also be discussed here.

... Types of negative dependencies Languages can express existential quantification in the scope of negation in a number of ways (cf. Chapters , ,  in this volume for discussion): ()

a. in the same way as the language expresses existential quantification elsewhere, by means of underspecified, multifunctional indefinites (e.g. bare wh-items); b. by means of negative polarity items (NPIs), used in all downward-entailing contexts, or in a subset thereof, and unable to express negation by themselves; c. by means of negative indefinites (NIs), items specialized for negative contexts and able to express negation by themselves, in Double Negation languages (where they cannot co-occur with the negative marker or with each other in a single-negation reading); d. by means of Negative Concord items (NCIs), also known in the literature as n-words. These items are (more or less) specialized for negative contexts and are able to express negation by themselves. In Negative Concord languages they can co-occur with the negative marker or with each other in a single-negation reading.

The distinction between NPIs (b) and NCIs (d) is controversial and is variously understood in the literature. In this chapter, I base the distinction on the property that NCIs have and NPIs lack: to occur in isolation and negate by themselves, for example in subject position or as short negative answers; this is exemplified by the contrast between Italian NCI nessuno ‘nobody’ and NPI alcuno ‘anybody’ in (–): ()

a. Nessuno è venuto nobody is come ‘Nobody has come’

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



  b. *Alcuno è venuto anybody is come

()

a. A: Chi è venuto? who is come A: ‘Who has come?’

B: Nessuno nobody B: ‘Nobody’

b. B: Chi è venuto? who is come

B: *Alcuno anybody

I also assume that NCIs undergo syntactic licensing by a negative operator (Negative Concord), whereas NPIs are subject to a different licensing mechanism, which is based on their pragmatic properties (see Penka : –, and Chapter  in this volume for discussion) and is available both in Double Negation and in Negative Concord languages.¹ In Negative Concord, multiple morpho-syntactic expressions of negation contribute a single logical operator to the semantic component. In Double Negation systems, instead, there is a one-to-one correspondence between the expression and the interpretation of negation. Diachronically, shifts between these two macro-systems are well documented in some branches, such as Germanic and Romance (cf. Willis, Lucas, and Breitbarth b). This raises a number of questions with a potentially relevant impact on theories of negation systems and of language change: what causes the shift from one system to the other? Is there a connection between changes affecting the negative marker and changes affecting the indefinites in its scope? In this section I critically review answers that have been proposed to these questions. I present examples of the evolution of Negative Concord and I discuss its connection with Jespersen’s Cycle. I review the various hypotheses that have been put forward for the emergence of Negative Concord: the hypothesis—on which most diachronic research has concentrated—according to which Negative Concord arises as an effect of the changing properties of the negative marker (..); the hypothesis according to which Negative Concord arises as due to changes in the internal syntax of the indefinite (..); the hypothesis according to which the interaction with Focus plays a role (..).

¹ A well-known problem for the assumption of a clear-cut distinction between NPIs and NCIs comes from the fact that some NCIs do show NPI uses, where they appear in non-negative contexts with a nonnegative meaning, witness Italian nessuno in (i): (i)

Mi chiedo se nessuno abbia poi contattato Gianni ‘I wonder whether anybody has eventually contacted Gianni’

(Rizzi : )

Cf. Milner () for early observations on similar facts in French, and Martins () for an overview of the situation in Romance. These phenomena have been interpreted as a consequence of the diachronic connection between NPIs and NCIs that will be explored in this chapter (cf. especially ..). However, no consensual analysis has emerged for their persistence at diachronically stable stages: cf. recent attempts in Chierchia (: –), Labelle and Espinal (), Gianollo (: ch. ).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



... Negative Concord as syntacticization of the licensing relation Rowlett (: ch. ), summarizing a number of observations in previous literature, states what he calls “Jespersen’s Generalization” in order to account for the persisting intuition that there is a connection between changes affecting the form of the negative marker and changes affecting the semantic and syntactic behavior of indefinites in the scope of negation. According to Jespersen’s Generalization, languages with a morphophonologically reduced negative marker exhibit Negative Concord, while languages with “fuller negatives” do not (Jespersen : –; Jespersen : ). That is, Jespersen’s Cycle, affecting the form of the sentential negative marker, is systematically connected with the presence or absence of Negative Concord. In generative terms, the morphophonological weakness of the negative marker has been taken to correlate with its status of syntactic head (X⁰), whereas “fuller negatives” have been taken to correspond to adverbial elements with a phrasal (XP) status, occupying Specifier positions (cf. Zanuttini  for a first cross-linguistic typology, and Chapter  in this volume). Further work helped refine this generalization, which incorrectly suggests that languages with XP negative markers will not have Negative Concord: this has been falsified by the existence of languages like Québecois French and Louisiana French Creole (Déprez ), as well as Bavarian (Weiß ). Hence, the generalization is not bidirectional. Moreover, a language that has a negative marker with head status may altogether lack indefinites that establish a specific syntactic relation with it (i.e. it may lack NCIs and use, instead, NPIs or underspecified indefinites to express existential quantification under negation, cf. (a) and (b)). A more accurate formulation of the generalization has been proposed by Zeijlstra (a, a, ), and states that if a language has Negative Concord its negative marker has head status.² ()

Negative heads are predicted not to be available in Double Negation languages. There is no Double Negation language that exhibits a negative marker that is a syntactic head.

This generalization leads to an important diachronic prediction: ()

Changes in the syntactic nature of the negative marker (from XP to X⁰) may trigger the creation of syntactic dependencies between negative items in the clause, leading to the birth of Negative Concord.

This step consists in the creation of NCIs and may be realized in different ways, depending on a number of factors: through the reanalysis of previous NIs of Double Negation languages, or through the reanalysis of NPIs, or through the creation of completely new ² The generalization in (4) has been slightly reformulated with respect to the one given by Zeijlstra (: ) by substituting ‘Double Negation languages’ for the original ‘non-Negative-Concord languages’, in consideration of the existence of languages that only have indefinites of the type in (a) and (b): these are non-Negative-Concord languages that may have a head-like negative marker.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

lexical items. The common property shared by all these outcomes is the establishment of a syntactic relation, conditioned by negation, between various elements in the clause. According to Zeijlstra (a), the establishment of this syntactic relation is triggered by the activation of a clausal NegP projection that happens when the negative marker is reanalyzed as a syntactic head. The syntactic relation is modeled as Agree (Multiple Agree in Zeijlstra a, a; binary Agree in Haegeman and Lohndal ) between a hierarchically superior [iNeg] negative operator and an uninterpretable formal feature [uNeg] on the agreeing indefinites (NCIs). The [uNeg] feature of NCIs may diachronically emerge in a number of ways, which will be reviewed in what follows. Clearly, an improved understanding of these diachronic processes can provide important insights into the theoretical debate on the status of the categories seen in (), since the diachronic relations between them are plausibly motivated by similarities and differences in their semantic and syntactic content. A case in which previous NIs are reanalyzed as NCIs is represented, for instance, by the few cases in which Romance continues the forms of Latin NIs: Romanian nimeni ‘nobody’ from Latin acc. neminem ‘nobody’; Old French nul ‘no’, and other similar forms derived from Latin nullus ‘no’ (Labelle and Espinal ; Gianollo ). Probably another case in point is the Hungarian sem-series of NCIs, which in Archaic Old Hungarian show signs of an NI behavior (Kiss ). An example of the reanalysis of previous NPI indefinite pronouns as NCIs is Italian veruno (Ramat ), from Latin vere ‘really, indeed’ and unus ‘one’: this indefinite is first found in Old Italian in NPI contexts and only later specializes in direct negation contexts. NCIs that have NPIs as their sources clearly show that, as we will see in more detail in section .., an NCI does not necessarily contain an originally negative morpheme. Welsh neb ‘nobody’, which contains a negative morpheme (< Indo-European negative morpheme *ne + wh-stem *kwo), is a complex case: an original NI was first reanalyzed as NPI (as the distribution in Middle Welsh shows) and finally as NCI (the Modern Welsh behavior), cf. Jäger (: –), Willis (a).³ The process leading from Latin to the Romance languages attests to the creation of completely new NCIs (with frequent parallels in the various daughter languages), once the prerequisites for Negative Concord are in place: this is the case, for instance, with Spanish ninguno ‘nobody’ and Italian niente ‘nothing’ (Martins ; Gianollo ; cf. sections .. and ..). Note that the prediction in () is not completely deterministic: as mentioned above, languages with an X⁰ negative marker may altogether lack indefinites specialized for negative contexts and may use underspecified indefinites (cf. (a)) or NPIs (cf. (b)) instead. Studies on early stages of Negative Concord in various language groups (in particular Slavonic and Romance) have often observed a fluid situation in the texts, characterized by ³ The reanalysis of NPIs is a frequent source for NIs of Double Negation languages as well (cf. e.g. the change from Middle High German dehein ‘any’ to Modern German kein ‘no’, Jäger ). I follow here Zeijlstra (a) in assuming that NIs do not establish a syntactic licensing relation with a clausal negative operator (since they are semantically, not formally, negative). If one, instead, adopts Penka’s () proposal that a syntactic Agree relation between the negative operator and the NIs exists also in Double Negation systems, these cases of reanalysis could as well be analyzed as due to the syntacticization of the licensing relation.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



remarkable oscillations between strict and non-strict Negative Concord. The earliest documents show that in some cases an indefinite preceding the inflected verb is the only marker of sentential negation (non-strict or asymmetrical constructions), in other cases it is accompanied by the sentential negative marker (strict or symmetrical constructions); cf. Willis, Lucas, and Breitbarth (a: –), van der Auwera and van Alsenoy () for a comparative overview. This, together with the existence of contemporary varieties, such as Catalan, displaying similar optionality of the negative marker with preverbal NCIs (Déprez et al. ), raises still unsolved issues for approaches like Zeijlstra’s (a), which attribute the difference between strict and non-strict Negative Concord to the featural specification of the negative marker. According to Zeijlstra (a), the negative marker is [iNeg] in non-strict systems and [uNeg] in strict ones (where the licensing of [uNeg] items is performed by an abstract [iNeg] operator). The [uNeg] / [iNeg] difference in the interpretability of the negative feature on the negative marker can be understood in diachronic terms as determining sub-stages of Jespersen’s Cycle, whereby the [uNeg] marker of strict varieties results from further weakening of a [iNeg] marker. Willis, Lucas, and Breitbarth (a: –) state that change typically leads from non-strict to strict systems (extension of contexts), and that counterdirectional developments (French, Welsh) are only observed in cases where the negative marker undergoes substitution. Solutions that have been proposed to account for ‘mixed’ systems such as Early Romance and Catalan include ambiguity analyses of the negative marker (Déprez et al. ) and attempts to resort to independent conditions acting at the clause level (Gianollo : –) and / or in the indefinite items. The latter have been formulated either in terms of feature composition of the indefinite item (Martins ; Labelle and Espinal ) or of its internal syntax (Déprez and Martineau ; Poletto : ch. ), cf. section ... A further aspect showing that the interdependence with the negative marker does not exhaust the dimensions of variation and change among Negative Concord systems concerns the difference between Negative Doubling (Concord applies between negative marker and NCI) and Negative Spread (Concord applies only between NCIs). Some languages show that the two can exist separately; varieties where only Negative Spread is possible are typically those, like French or Middle Low German (Breitbarth ), in which the negative marker underwent lexical substitution during Jespersen’s Cycle: Concord among indefinites is retained, but the new negative marker does not (yet) participate in the syntactic dependency relation.

... Negative Concord as an effect of the internal syntax of indefinites In .. we reviewed what Déprez () calls an “inside-out” approach to Negative Concord: the presence of Negative Concord, and the variation observed across Negative Concord languages, depend on the properties of the projection for sentential negation, NegP. Déprez (, , ) has proposed an alternative approach, an “outside-in” view according to which variation in Negative Concord systems depends on internal properties of indefinites entering the Concord relation. The Concord relation, in turn, may take various forms depending on the nature of the indefinites involved. The locus of

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

parametrization is thus shifted from properties of the clausal NegP (a macro-parameter) to properties of the nominal constituent (a number of micro-parameters).⁴ This approach has in principle important diachronic consequences, since the trigger to grammar change will be located in the internal semantic composition of the indefinite, which is mirrored by its syntactic structure and affects its syntactic position in the clause, instead of in the properties of NegP. Under this account, the relation with Jespersen’s Cycle is less prominent (although the generalization seen in () is preserved), and the attention shifts rather to the (micro)-variation observed across and within Negative Concord systems. This approach has the potential to provide insights as to the nature of ‘mixed’ systems, such as those observed in Early Slavonic and Early Romance and attested by a contemporary variety like Catalan. However, its application to diachronic data has been more limited and needs to be worked out in more detail in future research. Déprez and Martineau () and Déprez () provide a diachronic analysis of French rien ‘nothing’ and personne ‘nobody’ in the outside-in perspective: a connection is established between the development of these NCIs and the loss of bare nouns (connected with the evolution of number marking). In Middle French rien ‘thing’ and personne ‘person’ are still nominal and occur within the NP (a). The DP thus contains a null determiner, which is deemed responsible for the syntactic dependency with respect to the negative operator. In Standard Modern French (b), instead, NCIs are determiners, that is they occupy a higher position in the DP and are consequently subject to a different mechanism of Negative Concord involving resumptive quantification. ()

a. [DP D0 [NP personne ]] b. [DP personne [NP N0 ]]

Middle French Standard Modern French

The diachrony of French is particularly interesting in that it shows two stages of renewal in the expression of negative dependencies: the first Old French texts witness NCIs (nuls ‘nobody’, nient ‘nothing’, nul ‘no’) derived from Latin material, either by reanalysis of former NIs or by innovative grammaticalization processes (cf. () below). Later texts (starting in the fourteenth century with aucun ‘some/any’) witness a process of replacement through the new series persisting until today, formed by originally positive items (cf. ..– ..) like personne, rien, aucun (Déprez and Martineau ; Déprez ; Ingham ; Labelle and Espinal ). This shift is apparently independent of changes affecting the negative marker, although it may attest to a functionally parallel phenomenon of creation of originally emphatic elements to the detriment of plain ones, which involves the whole system of negation: also the negative marker ne starts to be systematically reinforced by pas, with an originally emphatic function (Eckardt : ch. ). A similar motivation, in terms of emphasis, for the new indefinites is plausible in view of their etymological source, as we will see in section ..

⁴ A similar change of perspective has been suggested also for the analysis of the negative marker: in the NegP-free approach proposed by Breitbarth () on the basis of Poletto’s () synchronic analysis, differences in the syntactic status of negative markers are attributed to their variable internal complexity, and not to the presence or absence of a clausal projection for negation.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



Also Old Italian has provided evidence for proposals that explain micro-variation in Negative Concord by resorting to the internal syntactic structure of the involved pronominal items. Poletto (: ch. ), in dealing with optional Negative Concord in Old Italian (Martins ; Stark : ch. ), proposes that optionality is in fact governed by the distribution of NCIs, which in turn is influenced by their internal structure. The NCI niente/neente ‘nothing’ obligatorily undergoes Concord when it is found in the same syntactic phase as the negative marker (CP-TP) or at the edge of the lower phase, whereas Concord is blocked if it is lower in the clause. The varying positions correspond to a variable analysis of the NCI’s internal structure. As an argument, niente is a complex item consisting of various projections: Negative, Existential, Classifier (ente ‘being’) and a null restrictor (a). As an adverb, it is reanalyzed as a single word (b), which can access higher adverbial positions. The version containing the null existential determiner, instead, has to remain low, in the c-command domain of the negative marker that licenses it. ()

a. [NegP ne [ExistP
[ClassP ente [RestrP
] . . . ]]] b. [NegP neente ]

A further phenomenon that has been accounted for by appealing to the complex internal structure and featural specification of the indefinite item is represented by NPI uses of NCIs in non-negative contexts, attested in some old and contemporary Romance varieties (Martins ): cf. the Old French example in (), where the indefinite nul occurs in the antecedent of a conditional (an NPI context) with the meaning ‘any’ (and also the correlative particle ne is not interpreted negatively): ()

Se il y a frere ne sereur qui soit encuses de nul meffait . . . ‘If there is a brother or a sister who is accused of any misdeed’ (Document from Amiens, year ; Ingham : )

These uses are difficult to reconcile with an analysis of NCIs in terms of [uNeg] features, and cannot always be traced back to an NPI-origin of the NCI (on which cf. ..), since sometimes the etymological source clearly contains a negative morpheme (as in the case of Old French nul < Latin NI nullus). Various studies have resorted to a dual feature system to account for the NPI/NCI ambiguity (Martins ; Jäger ; Labelle and Espinal ; Longobardi ; Gianollo ), adding a feature responsible for polarity (e.g. AFF(ective), σ (scalar), uFoc (focus)). Gianollo (: ch. ) has implemented the featural hypothesis in structural terms, proposing that Old Romance NCIs formed with the Latin negative focus particle nec ‘not even’ (e.g. Spanish ninguno ‘nobody’, Italian nessuno ‘nobody’, where nec is combined with unus ‘one’) were originally emphatic. Consequently, they had a more complex structure, containing a DP-internal Focus projection (a) that accounted for the productive NPI uses in a system like Chierchia’s (), where NPI-licensing is connected to Focus operators. Later on they have been reanalyzed as simple words (b) and maintain NPI uses only residually in the contemporary varieties. () a. [DP [FocP nec [NP unus[NP ]]]] b. [DP nessuno [NP]]

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

... The role of focus The role of focus, introduced at the end of .., has often been invoked in the study of the evolution of negative dependencies, but it is still lacking consensual formalization. Emphatic (or scalar) focus plays a role in the processes of pragmatic reinforcement not only of the negative marker, but also of indefinites in the scope of negation, revealing analogous functional pressures towards expressivity. Emphatic focus corresponds to the meaning of English even: it presupposes the establishment of a scale of values, and targets the extreme value on the scale, signaling that the assertion is surprising or exceptional in view of some contextual assumptions. When emphatic focus is combined with negation, the resulting assertion excludes any value on the scale, resulting in a strengthened negative assertion. In (), ‘a word’ is considered the likeliest thing to be said (the minimal unit of an act of saying); yet, the most likely alternative in the pool of considered possibilities is negated, yielding a surprising assertion (‘even not x’), which automatically entails the negation of all stronger (less likely) alternatives: ‘a sentence’, ‘a discourse’, etc. ()

He didn’t say even a word

This very powerful mechanism is exploited to achieve emphasis in the expression of negation, both in the reinforcement of negative markers (‘not a step’, ‘not a crumb’) and, as we will discuss, in the creation of emphatic indefinites of the form ‘not even one x’. As seen in (), the class of Romance nec-words (Gianollo ) is newly coined in Romance on the basis of a Latin construction containing a scalar focus particle (nec ‘not even’) and a minimal unit (the numeral ‘one’, minimizers like mica ‘crumb’, generalizers like ente ‘entity’). A Late Latin example is provided in (): () nec verbum contra fratrem responderunt unum pro ea not.even one: for her: word: against brother: answer: ‘they didn’t answer even a word in her defense against the brother’ (Aug. serm. .) NCIs formed this way are common cross-linguistically, as Haspelmath (: –) shows (cf. ..), and their original pragmatic motivation is clear. This is not to say that indefinites formed this way are always emphatic in the languages where they are found: historically, their emphatic value may fade away, according to a process that is understood as an ‘inflationary effect’ and is attributed to overuse in contexts that do not justify the assumption of emphatic focus (Dahl ). An analogous process of pragmatic strengthening and subsequent weakening leads to the creation of complex (discontinuous) negative markers like French ne . . . pas, with the function of plain sentential negation, out of an originally emphatic combination of a simple negative marker and an optional minimizer. The still open issue concerns the role of focus in the creation of syntactic dependencies. Simpson and Wu (), presenting a number of case studies that also comprise the grammaticalization of French pas, propose that an original, semantically motivated Focus projection (topping various types of constituents) may decay into a purely formal agreement shell, causing the ‘redundant’ expression of a grammatical category. Given the fact

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



that many indefinites used in the scope of negation are indeed emphatic (NPIs, NCIs formed with focus particle), one may push this idea further and argue that the grammatical expression of emphasis amounts to the creation of a Focus Concord dependency. At a further stage, similarly to what happens with Jespersen’s Cycle, inflationary effects would lead to the loss of the emphatic strengthening component, and to the reanalysis of the dependency as a purely formal one, in terms of [uNeg] features (cf. Gianollo  for an attempt in this direction). The fact that emphasis is important for the establishment of negative dependencies is also suggested by languages in which NPIs are created by attaching a focus particle to a plain indefinite. Hindi (Lahiri ) combines the focus particle bhii ‘also, even’ with various plain indefinites (e.g. ek ‘one’, koii ‘someone’), for example ek bhii ‘even one, any’. The NPI reading, emerging in downward-entailing contexts, is emphatic in virtue of its obligatorily focused status and of the scalar component contributed by the focus particle. In Modern Greek (Giannakidou ) there is a series of items where the same indefinite basic form has a non-emphatic and an emphatic version (e.g. kanenas ‘anybody’, KANENAS ‘nobody’; tipota ‘anything’, TIPOTA ‘nothing’, where capitals indicate emphatic accent). Both versions are grammatical under negation, but the emphatic forms are restricted to negative contexts, ‘without’-contexts, and ‘before’-contexts, whereas the nonemphatic forms have a broader distribution. Moreover, only the emphatic version can access positions linearly preceding their licensor (behaving like an NCI in strict Negative Concord). However, Giannakidou (, a) explicitly argues against a focus-based account of these facts. Therefore, the unification of the cases discussed above remains material for future research.

.. T    

.................................................................................................................................. In this section I review various ways in which an element is recruited to become part of the negation system of a language. I first deal with how pronouns and determiners become NCIs (..). I then move to items that are recruited from the descriptive lexicon and briefly deal also with the origin of minimizers and generalizers (..).

... NCIs with a pronominal origin .... Reanalysis of existing functional items One possible source for NCIs is the reanalysis of NIs in the shift from a Double Negation system to a Negative Concord one. We reviewed some of these cases in ... They are quite rare, possibly due to the more general cross-linguistic rarity of Double Negation systems. More frequently, NCIs derive from formerly more ‘positive’ items, in ways that will be discussed here and in .. (distinguishing, respectively, between a pronominal and a nominal origin). Another attested strategy is the creation of new items of the functional

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

lexicon: I will discuss one particularly productive strategy involving the combination of focus particles with a pronominal base (...). The shift from ‘positive’ to ‘negative’ has a systematic nature, in that it follows crosslinguistically recurrent steps of development: in the literature it is often discussed under the heading of “Argument cycle” (Ladusaw ) or “Quantifier Cycle” (Willis b; Willis, Lucas, and Breitbarth a, b; van der Auwera, Krasnoukhova, and Vossen forthcoming). I will first describe the process and then discuss which aspects of it may be understood as cyclical.⁵ The stages of the ‘Quantifier Cycle’ can be summarized as in (). ()

The Quantifier Cycle (i) plain indefinite / generic nominal expression > (ii) weak polarity item > (iii) strong polarity item > (iv) Negative Concord item or Negative Indefinite

Intermediate cases are attested, where indefinites undergo the cycle only partially, and do not reach stage (iii) or (iv).⁶ Concerning stage (i), the triggers to the change seem to be different in the case of elements already belonging to the functional lexicon and elements with a descriptive content. While the latter are usually unconstrained in their distribution across logical contexts, as is to be expected of elements of the descriptive lexicon, functional items (pronouns and determiners) acting as starting point of the process typically already display some contextual restrictions. This is the case, for instance, with Latin aliquis ‘some or other’, from which Romance indefinites such as Portuguese algum, Spanish algún, French aucun, etc. derive (through an intermediate reconstructed stage of combination with the cardinal numeral unus ‘one’: *alicunus). Latin aliquis has the meaning and the distribution of an epistemic indefinite (Gianollo : ch. ): it signals that the speaker is not able, or does not want to univocally identify a discourse referent. Epistemic indefinites are allowed in a set of non-assertive contexts where the inferences triggered by their epistemic semantic component are felicitous (Alonso-Ovalle and Menéndez-Benito ); being impossible in purely assertive contexts, they do not qualify as prototypically ‘positive’. The cycle affecting the Romance continuations of aliquis, which shows interesting parallels with what Hoeksema () observes for Dutch enig, yields different outcomes in the various languages. In Spanish and in Italian the continuations have an NPI distribution (stage ii). Spanish also retains an epistemic version, in addition to the NPI: epistemic and NPI uses are disambiguated by inversion of the determiner and the nominal restriction, which is obligatory in NPI uses. Also Portuguese retains the epistemic meaning, which is available when the indefinite precedes the nominal restriction (a). When, instead, inversion takes place and the indefinite follows its nominal restriction, Portuguese algum behaves as an NCI (Martins ⁵ A similar systematic process from ‘positive’ to more ‘negative’ is attested in the ‘Free-choice cycle’ (Haspelmath : –), which more frequently yields NPIs rather than NCIs, cf. Willis, Lucas, and Breitbarth a: –: one outcome of this process is Modern Greek tipote ‘anything’ < indefinite ti + adverb pote ‘ever’ (Horrocks ). ⁶ Jäger (), in her study of the diachrony of High German negation, has shown how NCIs can be reanalyzed as NIs, becoming this way even ‘more negative’ (semantically negative).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



), negating the sentence by itself when preverbal (b). In other words, inversion disambiguates between the negative and the non-negative uses. ()

Portuguese (from Martins ) a. Algum animal vive aqui some animal lives here ‘Some animal lives here’ b. Animal algum vive aqui animal some lives here ‘No animal lives here’

Also in Standard French aucun is an NCI. The reanalysis of the original epistemic indefinite into a weak NPI (from stage (i) to stage (ii)) can be traced back to Late Latin, where aliquis substitutes for other indefinites in contexts like the antecedent of conditionals and the scope of negation (Gianollo : ch. ). Later language-specific developments account for the differentiation of outcomes across Romance. Also the reanalysis of NPIs as NCIs may be understood as part of the Quantifier Cycle: in this case the elements representing the starting point of the change, as far as we can reconstruct it in the historical documentation, correspond to the stages (ii) or (iii) in (). According to a line of explanation going back to Ladusaw (), the developments described by the Quantifier Cycle are tightly connected to Jespersen’s Cycle, in that both originally positive items and NPIs are recruited as negation strengtheners. More in general, it is possible to recognize an analogous functional motivation for Jespersen’s Cycle and the Quantifier Cycle once they are interpreted as pragmatically and semantically driven chain shifts from ‘more emphatic’ to ‘less emphatic’. Traditionally, Jespersen’s Cycle has been understood as motivated by the weakening of the original negative marker, which prompts the recruitment of a reinforcer (pull-chain scenario). However this analysis encounters a number of theoretical and empirical problems (cf. Hopper and Traugott :  and Kiparsky and Condoravdi  for discussion), which led to the proposal of an alternative strengthening model: under this view, the trigger to the cycle is the mechanism of constant competition between newer, more expressive forms and previously available forms, which eventually recede under the pragmatic pressure of the competitors (push-chain scenario). This mechanism may be argued to apply to negative markers and indefinites in the scope of negation alike: the starting point of the cycle can be seen in the creation of an emphatic variant that eventually substitutes the older form in its function (respectively, the expression of sentential negation for the negative marker and the expression of narrow-scope existential quantification for the indefinites). This new form is, in turn, subject to ‘inflationary effects’ (..), which lead to the loss of emphasis and, potentially, to the start of a new cycle. This scenario connects in interesting ways to theories that motivate the existence of NPIs in virtue of their emphatic function, either in terms of domain widening or in terms of scalar alternatives (cf. Chapter  in this volume). The various analyses foresee felicity conditions for NPIs that ensure that they will be used only in contexts where they lead to an informationally stronger statement, that is in downward-entailing contexts.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

While approaches based on emphasis provide a plausible explanation for scalar NPIs, they do not apply as well to NPIs from a non-scalar source (the ‘referentially deficient’ items discussed by Giannakidou , e.g. emphatic tipota of Modern Greek seen in ..). Further diachronic research should ascertain whether the NPI-to-NCI shift seen in () predominantly applies to scalar NPIs. Studies that analyzed in detail the processes described above have often observed that the changing distribution of an item over time mirrors degrees of negativity in the various downward-entailing contexts, as determined by the logical inferences licensed in subsets of these contexts (anti-morphic and anti-additive; cf. Zwarts ). Alternatively, clines of expansion or restriction of contexts have been formulated on the basis of varieties of nonveridicality, which include, besides negative contexts, also questions, generic contexts, the antecedent of conditionals, the scope of modal verbs, etc. (Giannakidou ; Giannakidou and Quer ). The investigation is complicated by the fact that often system-internal effects (e.g. blocking by another indefinite) obscure the generalization underlying a given distribution. The development of [uNeg] features characterizing NCIs (cf. ..) has been attributed to the progressive restriction of licensing contexts for NPIs (Martins ; Zeijlstra a, a; Jäger , ; Biberauer and Roberts ). As mentioned in .., the ‘positive’ origin of some NCIs explains the fact that an NCI does not necessarily contain an originally negative morpheme.

.... Creation of new functional items A different diachronic phenomenon yielding NCIs consists in the creation of new elements of the functional lexicon, which are directly born as elements of Negative Concord. Resorting, once again, to the wealth of evidence provided by the diachrony of Romance, we observe that during the transition from Latin to Early Romance new indefinites emerge, which, as seen in .., are characterized by the combination of the negative focus particle nec ‘not even’ (neque is the fuller form continued by Romanian nici) and the cardinal numeral ‘one’. These nec-words (e.g. Portuguese nenhum, Spanish ninguno, Old French neuns, nesun, Romanian nici un, all meaning ‘nobody’) are formed similarly to the Slavonic ni-series < ni ‘not even’ + wh-indefinite (e.g. Russian nikto, Polish nikt, Bulgarian nikoj, all meaning ‘nobody’) that can be reconstructed for Common Slavonic (Willis b). Also Ancient Greek has an etymologically similar series of NCIs, built by adding oudé ‘not even’ to a number of bases (e.g. the numeral heĩs ‘one’ in oudeís ‘nobody’), cf. Denizot (). The Hungarian sem-series mentioned in .. has a parallel origin: the particle sem ‘not even’ originates from the univerbation of negation and an additive focus particle, similarly to Latin nec (< neque < negative *ne + conjuction/additive particle -que). Haspelmath (: –; –) discusses many cross-linguistic parallels for indefinites formed with a scalar focus particle ‘even’: some of them are NPIs (Modern Greek kanenas ‘anyone’ < kan ‘even’ + enas ‘one’; Dutch ook maar iets ‘anything’ < ook maar ‘even’ + iets ‘anything’), some are NCIs, and only a subclass contains a negative morpheme. The crucial ingredient appears to be the focus particle, which signals the presence of active alternatives in the interpretation. As discussed in .., alternatives on which ‘even’ operates are ordered on a scale, of which the indefinite base expresses an extreme point.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



The use of a focus particle to reinforce the indefinite base is remindful of functionally analogous processes that apply in Jespersen’s Cycle and similarly exploit the emphatic potential of scalar focus (cf. observations in Willis, Lucas, and Breitbarth a: ). In this case, under some analysis of Jespersen’s Cycle, besides the functional motivation, the process would also share the aspect of formal strengthening: this consists in the presence of a Focus shell, modifying the negative marker in the case of reinforced complex forms (Simpson and Wu : –; Breitbarth : –) or inserted in the DP in the case of nec-words and the like (cf. () and Gianollo : chs. –). Moreover, similarly to new negative markers emerging as the output of Jespersen’s Cycle, these NCIs, originally built as emphatic items, lose their strengthening function in the course of time, becoming plain NCIs. A similar cyclical shift from emphatic to plain is proposed by Chierchia (: –) as a correlate to his typology of NPIs, where he distinguishes between potentially emphatic NPIs (e.g. unstressed any) and obligatorily emphatic NPIs (e.g. stressed any), the latter activating scalar alternatives. Structurally, emphasis loss corresponds to structural reduction (loss of the Focus projection and development of uninterpretable formal features). The comparative history of Romance varieties shows particularly clearly how indefinites with an analogous syntactic behavior or even belonging to the same series may have heterogeneous etymological origins: Portuguese nenhum is a nec-word, algum derives instead from a former epistemic indefinite reanalyzed as NPI and later as NCI; Spanish ninguno is a nec-word, but nada ‘nothing’ derives from an item of the descriptive lexicon, Latin (rem) natam ‘born thing’. A similar situation is observed for the German series of NIs: while most items result from incorporation of a negative element (niemand ‘nobody’ < negative morpheme n- + ioman ‘anyone’; nimmer ‘never, no more’ < negative morpheme n- + immer ‘always’), the adjectival determiner kein originates from an NPI (Old High German dehein ‘any’) and does not contain a negative morpheme (Jäger ). A final issue I mention in closing this section concerns the debate on the directionality of the systematic developments surveyed here (Haspelmath ; Roberts and Roussou ; Biberauer and Roberts ; Hoeksema ; Jäger ). Traditionally unidirectionality from ‘positive’ to ‘negative’ has been assumed, on the basis of cases like those presented above. However, Jäger () has brought to attention cases where the development goes from more negative to more positive (e.g. Old High German NPI ioman ‘anyone’ to Modern German jemand ‘someone’), and has proposed a system of features and strategies of reanalysis to account for these developments in a unified way. Also cases where NCIs that derive from the reanalysis of NIs show NPI uses, as happens with the Romance continuations of Latin nullus (cf. ), attest to an apparently counterdirectional development (Labelle and Espinal ). The debate is still open on whether all these cases may be treated as unitary diachronic phenomena by resorting to a restricted featural system. Moreover, the evidence collected until now suggests that the change from ‘positive’ to ‘negative’ occurs more frequently than the change from ‘negative’ to ‘positive’.

... NCIs with a nominal origin, minimizers, and generalizers When NCIs originate from the grammaticalization of elements with an originally descriptive content, the starting point of the process is represented by nouns whose denotation

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

qualifies them to act as generalizers. Generalizers express a descriptive content that has the potential to apply to a broad set of individuals or entities. When grammaticalized as elements of the functional lexicon, they widen the domain of quantification (Zeijlstra : ) and are similar to free-choice items, in that they invite consideration of all possible scenarios, excluding all of them if combined with negation (Willis, Lucas, and Breitbarth a: ). NCIs originating this way are familiar from the history of Romance and often coexist with the source item in the descriptive lexicon: French personne ‘nobody’ < ‘person’; rien ‘nothing’ < ‘thing’; Catalan gens ‘no’ < ‘kind’; cap ‘no’ < ‘head’; res ‘nothing’ < ‘thing’; Spanish nada < ‘nothing’ < ‘born (thing)’, that is whatever exists, etc. Minimizers, denoting minimal quantities, are originally nouns that undergo grammaticalization, becoming adverbial elements that function as strengtheners of the negation marker (Batllori ): French pas < ‘step’, Italian mica < ‘crumb’, English a wink, Old High German drof < ‘drop’ (cf. Chapter  in this volume). In the case of both generalizers and minimizers, the first stages of the grammaticalization process witness the nominal item as a verbal argument. Typically, the syntactic position from which the grammaticalization begins is the direct object position, where the source nominals are originally selected by verbs with a compatible lexical semantics (e.g. ‘drink’ for ‘drop’). Only later, as the bleaching of the original lexical content progresses, they are found with verbs with which they are compatible only in the more general meaning of ‘generic being’/‘minimal quantity’ (Breitbarth : –). While generalizers typically develop into indefinites in the scope of negation, minimizers typically take on an adverbial function at the end of the grammaticalization process. Interestingly, minimizers can persist as bare nouns in languages otherwise prohibiting or strictly restricting the occurrence of bare singulars. For Modern Greek, Giannakidou (: –) proposes to treat bare singulars like leksi ‘word’ in () (where capitals indicate emphatic accent) as incorporated nominals. ()

Modern Greek Dhen ipe LEKSI oli mera not said word entire day ‘S/he didn’t say a word all day’

(Giannakidou : )

As concerns generalizers developing into NCIs, the most discussed examples in the literature come from the history of French, where the contemporary NCIs personne ‘nobody’ and rien ‘nothing’ derive from originally positive expressions with a general descriptive content. Déprez () summarizes the most important steps in the grammaticalization of personne and rien as can be reconstructed in the extant corpus. A first important sign of the decrease of descriptive content and lexical features is the loss of gender and number features: while early witnesses show that rien and personne have retained the feminine gender of the Latin source item, in the sixteenth century signs appear that feminine gender is lost, and the words appear with default masculine gender (something that, given the morphology of French, can only be detected if they are accompanied by agreeing determiners or adjectives); similarly, plural forms are lost for rien by the end of the sixteenth

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



century. The fact itself, however, that these words can occur with a determiner attests to their nominal behavior at that stage. Another important diagnostics concerns the change in modification possibilities undergone by these items with the progressive loss of nominal nature. We saw in .. how, according to Déprez, this correlates with a change in the internal structure of the indefinite. It has been variously proposed that this may also be connected to the loss of bare nouns in French, which triggered the reanalysis of personne and rien as determiners. The diachronic path evidences an intermediate stage where the distribution is context-dependent in terms of polarity sensitivity (similarly to what has been proposed for minimizers, Eckardt : ch. ; Eckardt ). Labelle and Espinal () refine this reconstruction in terms of chronology and motivations: they show that personne and rien become NPIs before the disappearance of bare nouns in the sixteenth century, and propose that the emergence of polarity sensitivity (which they formalize as the addition of a [+σ] feature, cf. ..) is an independent process, starting already in the twelfth century for rien (contemporaneous to the loss of its lexical source) and in the thirteenth to fourteenth century for personne. The first NCI readings are attested very early, in the thirteenth century, for rien, and between the sixteenth and the seventeenth century for personne.

.. C     

.................................................................................................................................. Considerable progress made by synchronic and diachronic research in recent years has led to an improved understanding of many phenomena of change in negative dependencies. The powerful explanatory generalizations concerning the connection between Jespersen’s Cycle and the changes affecting indefinites dependent on negation are among the most important aspects discussed in this chapter. Thanks to the comparison with Jespersen’s Cycle, we now understand better the nature of pragmatic pressures on the system of negation, and we can model their constitutive effects on grammar more precisely, in terms of strategies for the emphatic strengthening of negative utterances. Comprehensive typological and formally oriented descriptions of the histories of negation systems in some languages, and comparative accounts thereof, allow us to single out some developmental tendencies, such as the favored sources for the grammaticalization of NCIs, the semantic and syntactic mechanisms involved, and the steps through which grammaticalization processes develop in time. Future research will have to deal with a number of still open issues and emerging questions. From a descriptive perspective, we are still missing comprehensive accounts of the comparative history of entire branches or language families. On the basis of such descriptions, one could attempt to reconstruct the starting conditions and to identify local causes for differentiation. In this respect, we are still lacking a principled treatment of systemic effects: are there global properties, for example in the stock of quantificational items or in the syntax of quantification and negation, characterizing languages with neither

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

NIs nor NCIs? How does blocking across paradigms of indefinite items work exactly? Is the approach to Negative Concord based on the internal structure of NCIs able to provide answers to some unsolved issues, such as the NPI/NCI alternation observed with certain lexical items? Is it possible to better formalize the role of focus at the syntax-semantics interface and to capture, this way, important generalizations concerning the formation of negative dependencies? A number of issues that emerged in the discussion will certainly benefit from theory-driven empirical work on lesser-studied languages.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ......................................................................................................................

       ......................................................................................................................

 ˊ 

.. I

.................................................................................................................................. T objective of this chapter is to provide a survey of the current linguistic research regarding the role of pragmatics for regular changes in the form and meaning of negation.¹ The universally accepted, if often implicit reasoning, is that where change implies at least two forms in competition, the one that is emerging or declining should be characterized by markedness.² The marked option is understood as less frequent and more contextually constrained than the default expression.³ One contextual constraint is for the emerging or declining form to have a pragmatic value.⁴ This naturally follows from the various principles (de Saussure’s opposition (), Kiparsky’s Elsewhere condition (), Grice’s Maxim of manner (), Clark’s principle of Contrast ()) that suggest that an alternative way of saying things tends to be associated with a different interpretation. To what extent do changing negatives relate to a pragmatic value? What are the conditions under which this relationship obtains according to current research? Answering these two questions constitutes the purpose of this chapter. The discussion is organized in three main parts. The first reviews studies that posit a role for pragmatics in the change of negation on the basis of intuition. The alleged pragmatic values are rarely defined, and quantitative evidence of the behavior of items across time is generally not provided. Hard evidence is presented in a small number of current studies

¹ The role of pragmatics is relevant for other aspects of the change in the form and meaning of negation. A reviewer points out “in particular the role of the hearer-oriented Neg-first principle, which Jespersen himself invoked to motivate the successive shifts in the positioning and form of negation” (Horn a: ff.). For reasons of space and coherence, such further aspects are not explored in this chapter. ² For a general assessment of markedness and language change, see Elšik and Matras (). ³ Other factors often relevant to markedness such as morpho-syntactic complexity (Haspelmath ) do not seem to be crucial where negation is concerned. ⁴ For an example of semantic differentiation relating to clausal negation, see Montébran ().

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ´

surveyed in section .. They show that markers introducing new negative structures initially emerge in clauses that are explicitly discourse-old. Section . presents the debate about whether the pragmatic value of markers correlates with quantitative rates of use. In the final discussion section, I stress that the change of negation is constrained by a pragmatic evolution pathway (going from categorical explicit discourse-old contexts, to categorical discourse-old, to an increasingly lesser representation of pragmatic values). This applies to those (mostly clausal) negatives that introduce a new Jespersen Cycle stage, and the steps delineating this pathway may relate to usage frequency.

.. I 

.................................................................................................................................. Early diachronic studies have noted a relation between “affective meanings” and change in the expression of negation. This is epitomized by Meillet (), who proposes that there is a spiral of change such that a new item is added to the expression of clausal negation to make it more “intense,” and that the intensity is lost as the new multipartite expression is increasingly used. Co-opted by current grammaticalization theory (Hopper and Traugott : ), the model is used to explain both the replacement of the preverbal negator Neg (as in Latin, Mellet , and Greek, Kiparsky and Condoravdi , Chatzopoulou )⁵ by the joint use of Neg with a postverbal Neg. A rationale is thus offered for the recurring cross-linguistic change from Neg alone to its joint use with Neg to the sole use of Neg. Known as the Jespersen Cycle (Larrivée and Ingham ; Gelderen ; on the history of the term and some early proponents, see van der Auwera ), this change can be seen in an idealised illustration from the history of English: () a. b. c. d.

Ic ne seah I ne saugh noht I saw not I didn’t see

(Old English) (Middle English) (Early Modern English) (Contemporary)

Neg Neg + Neg Neg Neg

Stage  Stage  Stage  Stage 

Such a change is described as associating to pragmatics in a number of Indo-European languages (Kemenade ; Detges and Waltereit ; Roberts and Roussou ; Eckardt ; Kiparsky and Condoravdi ; Biberauer ; De Cuypere ; Gelderen ; Horn a; Huddlestone ; Gelderen ; Larrivée and Ingham ; Willis, Lucas, and Breitbarth ; van der Auwera and Gybels ; Zeijlstra ) and beyond (Lucas ; Devos and van der Auwera ; Willis, Lucas, and Breitbarth ; Vossen and van der Auwera ; van der Auwera and Vossen ) (for more languages and references, including to synchronic data, see Larrivée a). Often loosely referred to as emphasis, the pragmatic value concerned is rarely defined. Transferable diagnostics and quantitative assessments are thus not provided to allow

⁵ A not dissimilar case is found in the existential negation cycle uncovered by Croft (), which does however not seem to involve the systematic intervention of pragmatics (for a recent discussion, see Veselinova ). We therefore leave this further cycle to one side.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

      



verification of claims. One consequence is that a single notion is applied to different phenomena. Let us look at one recent study on the diachrony of Greek clausal negation. Chatzopoulou () revisits the renewal of the preverbal negation in a language that does not expand into a Neg stage. ἐβούλοντο () ταῦτα δ’ οὐκ tauta d’ u:k ebu:lonto these   want... ‘They didn’t want these’

Attic Greek (th–th c. )

σϵ θϵωρούμϵν () ουδέν udhén se theorúmen  . see... ‘we do not see you.’

Late Medieval (th c. )

βλέπω () Δϵν σϵ dhen se vlépo  . see... ‘I do not see you.’

 u:(k)

 (u)dhén Standard Modern Greek

 dhen

It is argued, in classical fashion (see Kiparsky and Condoravdi ), that the new negator is initially an “emphatic/intensified” negator that bleaches out with increased use before ending up as a standard negation, as assumed with NEG in English illustrated in (). “Emphasis” is conceived of as a scalar relation that contributes to make a stronger statement. However, how is the emphatic character of the new negation at each stage to be recognized in historical texts? The reader is not told.⁶ Also, no quantitative information is provided about the relevant currency of markers through time, which might indicate when the marker has stopped being “emphatic,” when it becomes the default marker, and whether these two situations are coextensive. So much for clausal negation. What about the development of Negative Concord Items (NCIs)? NCIs are items that express both an ontological category (like inanimate) and negation on their own (in e.g. a fragment answer such as Nothing to a question What did you do last night?). Their evolution is also claimed to involve emphasis. It is crosslinguistically frequent for items like nothing and never to start as weak NPIs in contexts such as conditionals () and questions; develop uses in strong polarity contexts such as the direct dependence on a without phrase () and clausal negation; before they acquire negative uses on their own; as the proportion of their polarity uses decreases.⁷ This can be illustrated by French rien ‘nothing’⁸ (on the evolution of this and other NCIs in French, ⁶ Some criteria are discussed in Chatzopoulou (: –) in the transition of udhén from neg-word to Neg. Such cases (illustrated in ()), discussed by Bayer (), slightly differ from the usual NPI to Neg. ⁷ This description however does not entail that the pattern is irreversible; Hansen (, ) cites cases in French where the item is an n-word from the outset, and subsequently acquires NPI readings; see also Larrivée (b). ⁸ The three examples are taken from the  Norman legal custom text Très ancien coutumier de Normandie.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ´

see Martineau and Déprez ; Kallel and Ingham ; Labelle and Espinal ; Kallel and Larrivée under evaluation). ()

. . . se il a riens . . . if he has anything

()

. . . il l’aura . . . sanz riens clamer he will have it without demanding anything

()

l’acheteur . . . n’aura rien en échange the buyer will have nothing in exchange

The move from weak to strong NPI is assumed by Giannakidou () to correlate with the integration of a scalar component equivalent to natural language even in the semantics of the item (see also Heim ). This converges with the cross-linguistic observation that in some languages (Lahiri ; Lee ; Gianollo ), a morphological element equivalent to even enters in the composition of NCIs. It tallies with the conception of negative polarity items as often⁹ scalar (Israel b; Chierchia ), although whether strong NPIs are more scalar¹⁰ remains to be assessed. The crucial change from NPI to NCI is empirically described in the light of emphasis by some studies, including recent work on Portuguese (Pinto ) and Spanish (Poole ). Pinto considers the rare Portuguese NCI nemigalha ‘nothing’ that is in competition with the more general nada. It is claimed that the NCI is emphatic, on the basis of its morphological composition, that comes from an overt negation nen with minimizer migalla ‘a crumb’, literally therefore ‘not a crumb’. No quantitative survey is offered for nemigalha, which disappears in the sixteenth century, and it is not shown whether bleaching is involved, or when that would have occurred. Gianollo in her Habilitation () analyses Latin and Romance items integrating the negative particle nec. Many descendants and competitors of Latin nemo ‘nobody’ do integrate an instantiation of focus particle ‘even’, and therefore must at some stage have been emphatic. This is understood in the tradition of alternatives produced by a Focus value. Occasional quantitative observations are provided (especially regarding the post-nominal use of the NCI that is arguably emphatic). Of course, the question that arises is whether all NCIs initially are similarly emphatic, and when and how they stop being so. After all, not all NCIs are documented as being “emphatic,” and while some claims have been made for the evolution of NCIs like French personne (Winters ; Eckardt ), the demonstrations are still to be provided. Similar attempts have been made in the generative tradition to use syntactic Focus¹¹ to account for the emerging concord value of NCIs (Simpson and Wu ; Watanabe ; Biberauer and Roberts ). The proposal is that an NCI disallowing concord with another negative can be introduced in a concord context if it is inside a Focus shell. The shell

⁹ But not always, see Duffley and Larrivée (). ¹⁰ To the extent that being “more” scalar makes sense at all, as pointed out by David Willis (personal communication), since intuitively, an item either does or does not evoke a scale. ¹¹ Focus being one of the ways in which emphasis is realized; see the discussion in Eckardt ().

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

      



prevents the concord relation, maintaining the grammaticality of the sequence. As the Focus bleaches out, concord is interpreted to be allowed. It seems true that changes in concord potential seem to relate to some “emphatic” value: this is the case in contemporary Finnish where a NCI normally found with a concording negatively-marked verb is being used in the vernacular without concord (Kotilainen ; for a similar case in Brazilian Portuguese, see Schwenter, Dickinson, and Lamberti ). In summary, a number of studies refer to an intuitive notion of “emphasis” to account for the evolution of negative markers. It is generally believed to apply to the initial uses of an item as a negative. Either a competitor is involved, or a morphological focus/scalar component is present, or a new stage of the Jespersen cycle is introduced. Clausal negators constituting a new stage of the Jespersen cycle seem to associate robustly to a pragmatic value, which is not always clearly the case with NCIs. The precise characterizations of the pragmatic value involved and the frequency of the relevant items are not generally provided; they are explored in some studies, which are discussed in sections . and ..

.. P 

.................................................................................................................................. Section . has bemoaned the use of a blanket notion of “emphasis” to describe the pragmatic import of negative markers. Not that markers never have an emphatic value. There are emphatic markers, resulting in a stronger statement than one with the default negative, and they can indeed be conceived of in scalar fashion. Witness the contemporary French examples below evoking that the speaker ‘slept not at all’ (see also Lucas and Willis ). ai pas dormi du tout. () a. J’ - have- - not sleep- at all b. J’ - c. J’ -

ai rien dormi. have- - nothing sleep- ai dormi rien du tout. have- - sleep- nothing at all

These cannot be subsequently hedged (), unlike the default, non-emphatic negative (). () % J’ai rien dormi. Peut-être un petit peu, mais pas beaucoup. ‘I slept not at all. Maybe a little, but not much.’ () J’ai pas dormi. Peut-être un petit peu, mais pas beaucoup. ‘I didn’t sleep. Maybe a little, but not much.’ Also, and crucially, they are not informationally constrained by the antecedent context. They can be used out of the blue, as an initial contribution to a dialogue or as a response to a question such as ‘What’s going on?’ or ‘How are you?’, which does not evoke sleeping. However, the antecedent context has been found to restrict the historical evolution of

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ´

emerging markers, especially in relation to a new stage of the Jespersen Cycle illustrated in () above (Schwenter ; Hansen and Visconti ; Kemenade ; Schwenter and Johnson ; Larrivée b, a, , ; Batllori ; Wallage ; Blaxter and Willis ; Squartini ). These restrictions correspond to the clause in which the negator is used being explicitly activated. Activated clauses present information that is accessible not only to the speaker, but also to the hearer because the information that they convey has been made available before in the exchange (Dryer ). Now, there are two ways to make information available. One is for the clause to have been used explicitly in the antecedent context.¹² This is the case for the Brazilian Portuguese postverbal não, which can double the preverbal marker when the clause in which it occurs has been explicitly evoked in the antecedent context according to Schwenter (). In the following, the postverbal negator não used by speaker B would be infelicitous if the clause had not been primed by A, and is thus not found in out of the blue statements. ()

A. – Você desligou o fogão, né? ‘You turned off the stove, right?’ B. – Nossa! Eu não desliguei o fogão não! ‘Damn! I didn’t turn it off!’ (Schwenter : , example (b))

A second is for the information conveyed by the clause to be inferred or accommodated, due to clausal relations or accommodating constructions, such as causal environment or focus configuration respectively. The diversity of activation sub-cases used in the literature are given in Table ., reproduced from from Blaxter and Willis (), where explicitly mentioned corresponds to explicit activation and forward-inferable, backward-inferable, and common knowledge to inferable activation.

Table .. Final typology of stage II and III negatives (Blaxter and Willis ) Form

explicitly mentioned

forwardinferable

backwardinferable

common knowledge

completely new

Old Norwegian ekki

felicitous

felicitous

felicitous

felicitous

infelicitous

Old French ne . . . pas/mie

felicitous

felicitous

felicitous

felicitous

infelicitous

Catalan no . . . pas

felicitous

felicitous

infelicitous

infelicitous

infelicitous

Italian non . . . mica

felicitous

felicitous

infelicitous

infelicitous

infelicitous

Brazilian Portuguese não . . . não

felicitous

felicitous

infelicitous

infelicitous

infelicitous

Brazilian Portuguese ø . . . não

felicitous

infelicitous

infelicitous

infelicitous

infelicitous

¹² A variant is for the clause to be referred to by an anaphoric device, such as anaphoric verbs, which in Old French define the distribution of preverbal clausal negative non (see Larrivée b).

       Discourse-old Explicit /Anaphoric activation



Discourse-new Accommodated /Inferred activation

No activation

 .. Pathway of evolution of new emerging negative markers

This table further suggests that there is an ordered evolution of activated markers. All markers occur in explicitly activated cases, and so this must be the context where new markers emerge.¹³ Markers are gradually used in less constrained activated contexts, before the activation requirement is relaxed and the marker occurs in discourse-new environments, such as initiative interventions and out-the-blue statements. This can be represented by the pathway in Figure .. This pathway may have an impact on the syntactic distribution of items, as it is generally assumed that activated negatives are a Main Clause Phenomenon (Poletto ), although that remains to be established. Another dimension that needs documentation is that of declining negatives: one could believe that disappearing clausal negatives are found exclusively in activated contexts, having gone through the pathway of evolution above in reverse. This is what is proposed for the loss of Neg in contemporary French (Hansen ), although the contextual restrictions documented on the basis of attestations by FonsecaGreber () do not resemble activation (Larrivée ), or strict emphasis. A near categorical discourse-old value does seem to attach to some of the medieval English declining clausal markers, as indicated by the discussion of Wallage in section .. Another illustration might be provided by the case highlighted by Pinto (). That paper raises the issue of NCIs. Sadly, emerging (or declining) NCIs have not (yet) been shown to be constrained by an activated value. There is no case to my knowledge of an equivalent of declining (or emerging) nothing or never attested in (explicitly) activated contexts only. This is a curious asymmetry that calls for an explanation. This section has shown that there is a consensus as to the pragmatic value that accompanies emerging clausal negatives introducing a new Jespersen Cycle stage, which initially occur exclusively in an explicitly activated clause. This also applies to rare configurations that present new concord patterns such as negative doubling (Schwenter and Johnson ), at a rate well below the % threshold (Larrivée b). Emphasis may attach to items containing a scalar/focus component, and to those competing with the default clausal negation, such as NCI used as the postverbal negator of an intransitive verb (of the type of (b) J’ai rien dormi, Bayer ). Do the steps on this pathway of evolution relate to particular frequencies of use? This issue is discussed in section ..

¹³ Hansen () notes that main Old French Neg pas and mie are not strictly categorically discourse-old, as .% of occurrences are clearly discourse-new. This is interpreted as a data issue by Larrivée (): it is plausible that the earliest available material is posterior to the initial emergence of Neg, when it would presumably have exclusively been explicitly activated.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ´

.. Q  

.................................................................................................................................. Current studies support the intuitive idea that a pragmatic value is associated with new patterns of negation. Under some conditions, negators are initially used in an explicitly activated clause that contains information accessible to the hearer because it has been used in the antecedent context. From there, they become used in implicitly activated contexts; and finally, in unactivated environments, where they join the ranks of pragmatically unremarkable negatives. One question that this pathway of evolution raises is under what conditions an item progresses from one step to the next. Remember that the intuitive grammaticalization approach proposes that increased use leads a categorically pragmatically charged expression to have no particular pragmatic role. Is there an identifiable rate of use at which a marker ceases to be exclusively explicitly activated, or categorically pragmatically charged?¹⁴ Answers to these questions are provided in Larrivée (b). The paper is concerned with verifying whether negative doubling of the clausal negator with a NCI (as in the Rolling Stones’ I can’t get no satisfaction) is pragmatically charged or not (regardless of the construction’s register). In order to do so, a comparison is made between the activated status of doubling pertaining to two types of contemporary French varieties. The normative tradition of European French actively discourages negative doubling; it is an unexceptional feature of Quebec French. In corpora of vernacular exchanges, negative doubling is found to be rare in European French (.% of NCIs are involved in doubling in the Parisian corpus CFPP, .% in the France French corpus Clapi, and .% in the Belgian French corpus Valibel); it is not infrequent in Quebec French (at a rate of % in the CFPQ corpus). The prediction made is that below % of uses, categorical activation should be the case. It is supported by European French data, where negative doubling at well below % has a categorical pragmatic value (if not a categorical explicit discourse-old one). And by the Quebec French data where negative doubling, at well over %, has no characterized pragmatic value. The % mark is also found to associate with the categorical activation of in situ interrogatives (Larrivée ). A contrasting view is put forward by Phillip Wallage. His  article provides an analysis of the activated value of clausal negators in the history of English. English had the three stages of the Jespersen Cycle in the three periods examined (–, –, –), the preverbal Neg represented by the sole use of ne, the embracing Neg and Neg ne . . . not, and the use of Neg not on its own, as in () above. The move from stage  with ne alone to the concording stage  ne . . . not is found to correlate with a discourse-old value. Table . adapts the numbers from Wallage ().¹⁵ While the expression ne . . . not that is dominant from the intermediary period is less and less associated to a particular pragmatic value, the declining use of ne on its own is

¹⁴ For a quantitative model of the evolution of negation, see Vulkanovic (). ¹⁵ I’m leaving aside the category “counterfactual,” since it contains only twelve occurrences over the three periods, and more importantly it is not assigned a discourse value.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

      



Table .. Discourse-old and discourse new uses of ne and ne . . . not in the history of English –

–

–

ne

ne . . . not

ne

ne . . . not

ne

ne . . . not

Discourse-old

.% ()

% ()

.% ()

% ()



% ()

Discourse-new

.% ()

% ()

.% ()

% ()

% ()

% ()

()

()

()

()

()

Totals (numbers) ()

increasingly used in discourse-new contexts (whereas curiously, the option of not on its own is evenly distributed in old and new clauses). So while it is discourse-new rather than discourse-old values that are categorically associating to the minority variable, it is a straight case of the rare marker (less than .% of the total of  clausal negators) relating to an informational pragmatic value. But that is not the view put forward by Wallage. He argues that proportions do not represent language change adequately, because change follows a logistic curve measuring probabilities. A logistic regression analysis is applied to the pragmatic value of ne alone and ne . . . not, and he proposes that the rate is constant across values for the investigated period. Thus, although the expression ne . . . not becomes rather frequent (representing .% of all  occurrences of the three configurations in the intermediate period), it would not entirely lose its pragmatic value. Nonetheless, in terms of proportions, the expression ne . . . not does go from the nearly categorical pragmatic value of discourse-old to the absence of categorical value as it becomes more current. Another case looked at by Wallage is the competition between a clausal negator with an indefinite determiner (He didn’t see any rainbow) and a negative determiner (He saw no rainbow). The numbers for the present-day period are compared to those of – (Table . is adapted from Wallage : ).¹⁶ The not . . . any configuration clearly replaces no as the dominant strategy. In the earliest period, when it is marked, not . . . any has a dominant discourse-old value at nearly % of occurrences that is lost when it becomes the default option. When no goes from the default to the marked status, it associates with discourse-new clauses at well over %.¹⁷ In both cases, we have the expected progression: a marked expression tends to associate to a pragmatic discourse value, whether it be emerging or declining. More data would be needed in order to know if there was a time when the expression was exclusively explicitly activated (or in the case of ne and determiner no, only discourse-new), as the data reviewed ¹⁶ For more quantitative data on the alternation between not . . . any vs. no, see the discussion in Burnett, Koopman, and Tagliamonte (in press). Quantitative studies of the alternation do not to my knowledge include consideration of the clausal negative plus simple indefinite (He didn’t see a rainbow). ¹⁷ This may be why, as noted by a reviewer, no has been claimed to be more “emphatic” than its unmarked alternative not . . . any (Bolinger ).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ´

Table .. Discourse-old and discourse-new uses of not . . . any and no in the history of English –

Present-day

Not . . . any

. . . no

Not . . . any

. . . no

Discourse-old

.% ()

.% ()

.% ()

.% ()

Discourse-new

.% ()

.% ()

.% ()

.% ()

Totals (numbers)

()

()

()

()

in section . suggests. It remains to be seen whether a categorical pragmatic value generally correlates with a rate of use beyond what has been found for negative doubling. In any case, the proposal from Wallage that the categorical pragmatic value is lost only as a result of the expression accessing the default status is not supported (see also Hansen ): not any is clearly the default option as compared to no in the present-day versus the – period; yet, it is equally not categorical and predominantly discourse-old in both periods. Therefore, the loss of the categorical explicit activation significantly precedes the change of not . . . any to a default.

.. F 

.................................................................................................................................. Diachronic evolution sees new markers of negation display a pragmatic value. This is well established for new clausal negators (and possibly for NCIs) that introduce a new stage of the Jespersen Cycle: they are typically used in explicitly discourse-old clauses. A value of emphasis is recurrently suggested for NCIs such as nothing when they are synonymous with an existing competitor (Portuguese nemigalha vs. nada), contain a focus/scalar component (Romance expressions comprising nec) or enter in a new concord pattern (competing with clausal negator, as in (b)). It is unclear whether declining markers regularly associate with a pragmatic value as they disappear, although as it is about to disappear, English ne is categorically discourse-new. Discourse-old information is defined in terms of activation (Dryer ), which represents information accessible to the hearer, typically because it has been evoked in the antecedent context. A pathway of evolution can be identified such that new clausal markers introducing a new stage are initially explicitly activated, as in example (). They evolve by acquiring uses in implicitly activated clauses, before being also found in discourse-new contexts (see Figure .). This pattern of evolution is potentially associated with rates of use. One proposal is that explicit activation is found for clausal negative markers that represent less than % of all clausal negators. A general question that percolates from the proposed pathway aggregating documented results is why marked negatives should be sensitive to informational distinctions, not only

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

      



for change, but also for synchronic variation across languages. Another question is why this sensitivity should be related to discourse-old information. To put it in common parlance, why are new clausal negators initially used in sentences that echo a previous sentence? A possible explanation might be constructed from the observation that distribution in explicitly activated contexts relates not only to marked negation, but also to varying and changing expressions of interrogation (such as in situ questions, Rosemeyer ; Larrivée ) and possibly Verum Focus (as with English do support, Wallage ; see studies on various polarity particles in Romance, Batllori , Batllori and Hernanz , Del Gobbo, Munaro, and Poletto , Remberger ; on echoic responses, see Rosemeyer and Schwenter  and references therein). These all concern the actualization of the proposition they range over, and it seems reasonable that the proposition the actualization of which is discussed is explicitly stated, at least in the initial stages of development of the new expressions. This would explain why the pragmatic value of activation for rare veridical markers (Giannakidou ), such as negation, interrogation, and Verum Focus, occurs and is learned despite the paucity of input, and why it is stable across languages. We would therefore be dealing with a cognitive fact, rather than some version of the conversational principle of extravagance (Haspelmath  contra Chatzoupoulou ). Emotives of all sorts undergo a loss of noteworthiness through increased usage (Fabulous!, Poppycock!, Knave!) that leads to their replacement (Dahl ). But they are not contextually constrained in the way veridical markers are. This is a further reason to separate emphasis from discourse-old. The impact of pragmatics on the evolution of negation and other veridical markers raises a number of questions for future research. More empirical cross-linguistic work is needed concerning the quantification of the discourse-old value of emerging clausal negators, whether any NCIs would undergo the same process,¹⁸ whether there are either clausal negators or NCIs that this process does not affect, whether the process affects declining items (as it does categorically discourse-new declining English ne), and whether only veridical grammatical categories are involved. If the pathway articulated here is the case for a variety of veridical markers, this would suggest that a cognitive regularity is at play, and this should therefore be testable experimentally. What shape these experiments could take is not something that this author can define at present, so removed is their prospect. I have no doubt however that the creativity of future pragmaticists will also express itself in that domain.

A

.................................................................................................................................. I am grateful to Katerina Chatzopoulou, Chiara Gianollo, Maj-Britt Mosegaard Hansen, Ans van Kemenade, David Willis, and the reviewers for their comments on an earlier incarnation of this chapter. They are in no way responsible for the infelicities that it may still contain.

¹⁸ On the link between the presumed origin of negative markers and their syntactic and interpretative properties, see Poletto ().

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ........................................................................................................................

EMERGENCE AND ACQUISITION OF NEGATION ........................................................................................................................

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ......................................................................................................................

      -             ......................................................................................................................

 ,  ,   . ¨

.. I

.................................................................................................................................. I the philosophical literature, negation has been firmly tied to natural language or equivalent symbolic systems (e.g. Horn a; Bermúdez ). So far, there is no evidence that any nonhuman animal species might possess a communication system that relies on symbolic representations. As a consequence, animals might be ‘out of the game’ when it comes to negation. In this chapter we want to discuss findings from animal cognition research that suggest that animals think and reason in ways that resemble reasoning based on negation. In the main part of the chapter, we offer a more detailed review of some particularly promising findings from the literature on individual and social reasoning. While intriguing, we conclude that none of these provides definite evidence for reasoning based on propositional negation. In the third part of the chapter, we suggest ways to approach the question of propositional negation in animals empirically. Finally, we sketch out some ideas about the role of negation in the evolution of propositional thought more generally.

.. F  

.................................................................................................................................. In this chapter, we follow Bermúdez (, ) and distinguish between proto-negation and negation proper (hereafter negation). The core of this distinction is that while protonegation operates on at least two discrete representations that are contraries, negation

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ,  ,   . ¨ 

operates on a single representation (a proposition) that is evaluated as being the case or not. In Fregean terms (), negation operates on intensions, that is on propositions that represent the specific aspects of the world. Negating such propositions implies that whatever the intensional aspect is, is not the case. Proto-negation, on the other hand, operates on mutually exclusive extensions, that is, representations of certain states of affairs in the world. These representations can be thought of, to some degree, as pictorial representations of a given scene. As a consequence, a certain state of affairs is negated by finding out that the other member of the contrary pair is the case. Proto-negation nevertheless implies structured thoughts with elements that may reappear in different thoughts.

... Proto-negation Proto-negation is thought to be based on contrary pairs, such as ‘presence–absence’ or ‘danger–safety’. Let us illustrate the distinction between proto-negation and negation with an example. In a study by Call (), great apes were faced with two potential hiding locations of a food reward. In one condition, they were shown that one of the hiding locations, say location A, was empty. When making a choice, apes chose location B and therefore presumably reasoned that if the reward is not in A, it must be in B. On a protonegation account, rather than negating the presence of a desired food reward in A, individuals might have arrived at the conclusion by reasoning based on mutually exclusive states, namely that the reward is either present (absent) in A or present (absent) in B. When seeing that the reward is absent from A, apes concluded that the reward is present in B. The consequence of negating one state of affairs is thinking that a different state of affairs is the case. Let us elaborate: The straightforward way to express this inference in a formal way would be the disjunctive syllogism. ‘P or Q’, ‘not-P’ therefore ‘Q’, with P being ‘food present in A’ and Q being ‘food present in B’. This formulation, however, implies the ability to negate the proposition expressed by P, which should not be possible on a proto-negation account. Bermúdez () describes proto-negation as a form of predicate negation. Instead of negating the entire proposition P, only the predicate ‘present’ is negated. A negative predicate, he argues, can be substituted by a different predicate that has the same extension. In the above example, it would mean concluding ‘absent’ instead of ‘not present’. Both describe the same situation but absent is not constructed from present and therefore does not require an intension-preserving representation of present. As a consequence, protonegation as an operation allows an individual to switch between the elements of a contrary pair. Reasoning based on proto-negation therefore requires discrete contrary pairs (or sets, each with a limited number of elements). This begs the question of how animals come to form contrary pairs. Contrary pairs need not be mutually exclusive from a logical perspective but they should be mutually exclusive based on the experience of a given individual. If, for example, an animal experiences that two sorts of fruit are never available at the same time, the animal might come to form a contrary pair allowing it to infer the absence of one kind of food based on seeing the other kind. However, contrary pairs might be formed inferentially based on knowledge of spatio-temporal or object–object relations (Völter and Call ). For example, if an animal has some understanding of object solidity, it might

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



come to form the contrary pair that a board lying on the ground and an object being hidden under the board are mutually exclusive (see e.g. Call ). The way that contrary pairs are formed determines the range and flexibility of how an individual may be able to use protonegation, but it does not necessarily alter the reasoning process itself. To sum up, the important distinction is that conclusions based on proto-negation are limited to discrete states of affairs depending on the number and types of contrary pairs an individual is sensitive to. Which contrary pairs an individual is sensitive to depends to a large part on direct experience.

... Negation Whereas proto-negation operates based on contrary pairs, negation, on the other hand, operates on negating intensional aspects of a given state of affairs. As a consequence, it allows individuals to conclude a wider range of potential alternatives. In our example above, from the fact that the food is not in A could also follow that the experimenter ate the reward or that an invisible goblin stole it. These alternatives are certainly less plausible, but nevertheless possible. Negation only implies that one state of affairs is not the case. From a representational point of view, being able to negate a state of affairs requires an individual to form a representation that captures the intension of a proposition with which it could be described. To illustrate this point, think of an animal that represents states of affairs in a purely pictorial way. Such an individual would not be able to differentiate between two distinct states of affairs that have the same pictorial extension, say an empty box. For example, to differentiate between the thoughts that the banana is not in the box and that the grape is not in the box, they would need to be structured in a way that they capture a certain aspect (or perspective) on a visual scene. Among others, Bermúdez (, ) argues that the necessary structure is only provided by representations that are based on a system analogous to human language. By convention, symbols allow the representation of intensional aspects of experience. This is argued to be a necessary prerequisite to negate this very aspect. We do not necessarily commit to the claim that natural language is a necessary prerequisite for negation, but we do acknowledge that thoughts need to be structured so that they allow for representing intensional aspects.

... What do you need negation for? In the following, we explore the evolutionary relevance of negation by asking how the capacity for negation might impact an organism’s cognition and as a consequence its behavior. In each case, we start with considering empirical findings in animals that seem to be related to negation. We discuss what animals are capable of doing and if/how negation proper would potentially alter this ability.

.... Individual reasoning In their natural environment animals often face negative information. Negative information refers here to the perceptual absence of a stimulus or cue, that is a situation in which an

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ,  ,   . ¨ 

individual currently cannot perceive a stimulus. Perceptual absence includes occlusion events and situations in which there is actual evidence about the absence of a stimulus (henceforth: explicit absence). A central question is whether animals can distinguish different types of negative information, that is between occlusion events (i.e. absence of evidence about the current state of a stimulus) and explicit absence. Multiple lines of cognitive research are relevant to this question including studies on object permanence, inference by exclusion, associative learning, and information-seeking. Object permanence can be seen as a basic form of inference about negative information that helps to discriminate between occlusion events and explicit absence. Object permanence was first studied by Piaget () in human children. In the benchmark test of object permanence, the subject witnesses how the experimenter (E) hides the reward under a small opaque box, the displacement device. E then moves the displacement device under one of multiple opaque cups that serve as potential hiding locations. When the displacement device reappears from the visited cup, E shows the subject that the displacement device is now empty. Piaget found that children between  and  months of age can locate covered objects after so-called invisible displacements. Moreover, developmental research provided evidence for an incremental understanding of occlusion events during the first year of life (based on analyses of infants’ looking times; e.g. Aguiar and Baillargeon ; Luo and Baillargeon ). Following the Piagetian tradition, comparative research examined object permanence abilities in a large number of species (see Jaakkola  for a recent, comprehensive review). However, due to methodological reasons, only great apes (e.g. chimpanzees, Collier-Baker and Suddendorf ) and two parrot species (Goffin’s cockatoo (Cacatua goffini), Auersperg et al. ; grey parrot (Psittacus erithacus), Pepperberg, Willner, and Gravitz ) have so far provided conclusive evidence that they can locate objects after invisible displacements. A complementary way to study occlusion events (without the requirement to track displaced objects) can be found in the literature on information seeking. Information seeking studies typically address whether subjects seek additional information selectively when they are missing a relevant piece of information (Call and Carpenter ). For instance, in the most prominent paradigm, subjects are presented with a number of horizontal, opaque tubes that serve as hiding place for a food reward. In the critical experimental manipulation, subjects either witness the baiting of one of the tubes or not. A number of primate species including great apes, rhesus macaques (Macaca mulatta), and brown capuchin monkeys (Sapajus apella) seek information before they choose by looking into the opaque containers in particular when they could not see the baiting beforehand (Call ; Call and Carpenter ; Hampton, Zivin, and Murray ; Marsh and MacDonald ; Vining and Marsh ). Great apes also seek information about the location of a relevant non-food item such as a functional tool (Bohn et al. ). Information seeking was also found in the context of a delayed matching-to-sample task in which subjects either got to see the sample stimulus or not. Rhesus macaques and brown capuchin monkeys, unlike pigeons (Roberts et al. ), sought information about the sample stimulus when they could not see it previously (Beran and Smith ). Whether selective information seeking is supported by meta-cognitive capacities is subject to ongoing debate (Crystal and Foote ). Irrespective of this debate, the findings

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



suggest that there are a number of primate species that compensate absent evidence about the location of objects by seeking additional information. Another line of research relevant to the question of how animals represent negative information has been conducted with rats (Rattus norvegicus). In these studies, rats learned that a stimulus (e.g. a light) predicted a food reward. Following the conditioning procedure, rats were presented with negative information, either with an ambiguous occlusion event (e.g. a covered light bulb) or with an explicitly absent stimulus (a visible but unlit light bulb; Blaisdell et al.; Fast and Blaisdell ; Fast, Biedermann, and Blaisdell ). Rats discriminated between the covered and explicitly absent stimulus and treated the ambiguous covered light bulb more like the present stimulus (the lit light bulb) than the explicitly absent stimulus (unlit light bulb). In a related study, rats distinguished between an occluded reward (a covered drinking receptacle) and an explicitly absent reward in a Pavlovian extinction paradigm (Waldman et al. ). Rats again first learned that a light predicted a reward. Then they received an extinction phase in which the light was not paired with the reward any more, either with a covered drinking receptacle (Cover condition) or with an accessible drinking receptacle that did not dispense the sucrose reward (No cover condition). That is, in the No cover condition the reward was explicitly absent whereas in the Cover condition the presence or absence of the reward was ambiguous. In the test after this extinction phase, the cover was removed. Rats’ expectations for sucrose when they saw the light were higher in the ambiguous Cover condition compared to the No cover condition, in which the reward was explicitly absent during the extinction phase. Explanations based on associative learning theory include the so-called renewal effect, which basically states that extinction is context specific (Dwyer and Waldmann ). Accordingly, the introduction of the occluder constitutes a change in context which would explain the recovery of a conditioned response toward the occluded option. Given that it remains ambiguous from this hypothesis what constitutes a sufficient change in context, explanations based on the renewal effect are difficult to falsify. Another associative learning explanation suggests that explicit absence of the stimulus (e.g. the unlit light bulb) acquires associative value (Dwyer and Burgess ). When the light bulb is covered this should lead to a general effect on performance (given that the unlit bulb is not visible any more). In contrast, a recent study suggests that when presented with the covered light bulb rats’ behavior is governed by retrieved representation of the activated light and not by the fact the rats could not see the unlit light bulb (Fast, Biedermann, and Blaisdell ). Non-representational, associative learning based explanations therefore seem to be insufficient to account for rats’ performance in these tasks. Discounting these alternative explanations, however, does not mean that rats engage in negation proper when solving these tasks. Finally, the literature on inference by exclusion suggests that at least great apes, capuchin monkey, and grey parrots are able to infer the location of a reward by discounting alternatives (for reviews see Schloegl, Bugnyar, and Aust ; Völter and Call ). Typically, subjects are presented with visual or auditory information about the absence of a food reward in a two-choice task. For example, the subject might be presented with a food reward but it cannot witness under which cup the reward is hidden. In the critical condition, subjects get information about the baiting status of the empty cup, either by showing the content of the empty cup to them or by shaking the empty cup (e.g. Call ).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ,  ,   . ¨ 

While many species can solve the visual version of this task fewer species have been found to solve the auditory version of the task (Völter and Call ). In the latter auditory condition, the absence of a rattling sound signals the absence of the reward. Great apes (Call ; Hill, Collier-Baker, and Suddendorf ), capuchin monkeys (Sabbatini and Visalberghi ), grey parrots (Schloegl et al. ), and domestic pigs (Nawroth and von Borell ) were found to locate the food even when only the empty cup is shaken. Some of the great apes—gorillas (Gorilla gorilla) and bonobos (Pan paniscus)—and capuchin monkeys even adapted their behavior to the way the silent cup was manipulated by the experimenter: they avoided a silent-shaken cup but not a silent-stirred cup (that should not produce a rattling sound even if it contained a piece of food). In sum, convergent lines of evidence suggest that at least some nonhuman animals differentiate between different types of negative information. Specifically they discriminate between occlusion events and explicitly absent stimuli. However, all the reviewed examples do not demand negation proper. In most of the test situations, individuals need to consider specific alternatives suggesting that proto-negation might be sufficient to explain their performance. However, proto-negation with present–absent contrary pairs alone might not be a satisfying explanation because it does not explain how individuals discriminate between occluded and explicitly absent stimuli (Blaisdell et al. ) or between a silentshaken and a silent-stirred cup (Call ; Sabbatini and Visalberghi ). It is hard to falsify such proto-negation explanations because one could easily make up more specific contrary pairs that would explain their performance, such as occluded–absent, or shaken– stirred contrary pairs.

.... Social reasoning Flexible predictions about how others are likely to behave in the future depend, to some degree, on an assessment of the information they have been exposed to in the past (or not). In what follows, we will discuss the relevance of negation for social reasoning in competitive contexts as well as in communication. In order to do so, we assume that, at least some animal species, represent mental states of others. This assumption seems to be warranted in the light of the available empirical evidence (see e.g. Krupenye et al.  as well as Call and Tomasello , Whiten  for reviews), however, some scholars think that mental state ascription is not necessary to explain these findings (Heyes ). This discussion is relevant to the discussion of negation in social reasoning as it decides whether one thinks that the types of representations that (proto-) negation operates on are representations of directly perceivable events in the world or the mental states resulting from them. In any case, the question of what type of reasoning process might be used depends, at least according to Bermúdez (), not on the content of the representation but on its structure (propositional or not). A fair number of studies have looked at how animals strategically exploit others’ access to certain information in competitive contexts (e.g. Bray, Krupenye, and Hare ; Bugnyar, Reber, and Buckner ; Dally, Emery, and Clayton ; Flombaum and Santos  Schmelz, Call, and Tomasello ). For example, great apes have been found to avoid food that a more dominant individual can see or has seen in the past (Hare et al.; Hare, Call, and Tomasello ). Similarly, great apes avoid looking for food in places that would attract a competitor’s attention because accessing them makes noise (Melis, Call, and

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



Tomasello ). Let us describe a fairly recent version of this type of study. Karg and colleagues () conducted a study with chimpanzees in which the experimenter and the participant faced each other, separated by a mesh panel and a small opaque barrier and flanked by two potential hiding locations. These locations were elongated boxes with either an opaque or a transparent lid. Food items hidden in these boxes were accessible from the experimenter’s as well as the participant’s side. In the beginning of a trial, the experimenter would open the lids towards the participant to let them know which one was transparent and which opaque. Then she placed food items in both boxes and waited. The following interaction was structured so that the experimenter would pull away food items whenever she saw that the participant would reach for them. This was only the case when the participant reached into the box with the transparent lid. Results showed that apes reached into the box with the opaque lid more often than into the box with the transparent lid. On a generous reading, this implies that subjects represent a negated version of their competitor’s epistemic state (e.g. ‘does not see X’) and then choose among the behavioral options available to them which would not lead to consequences that are in conflict with this representation. As mentioned previously, this suggests a wide variety of potential alternatives. In this study, and also the ones cited earlier, individuals made binary choices (e.g. ‘reach for left food or right food’). This allows for a, presumably more parsimonious, explanation in that participants represented the two epistemic states held by the experimenter after looking at either the transparent or the opaque lid as a mutually exclusive contrary pair (‘sees content of left box or sees content of right box’). By seeing the position of the open lid and inferring that the experimenter would see the content of the box, participants could directly deduce that she would not see the content of the other box (as these two states were mutually exclusive). The subtle consequence of operating on a negated epistemic state of the experimenter (as opposed to a contrary pair) would be the availability of potential alternative strategies in case the primary option becomes unavailable. If, for example, only the box with the transparent lid were available, the same target condition (experimenter does not see content) would also be reached if the experimenter were for example to be distracted or tricked into looking elsewhere. Operating on contrary pairs would make this option not immediately available, as there has been no prior experience that the two states ‘experimenter looks at stone I threw to the left’ and ‘experimenter sees content of the box’ are mutually exclusive. While new contrary pairs like this could certainly be formed, they would not be immediately available. We hope that this example illustrates that using negation in social decision making increases the range of available strategies. Communication differs from competitive social reasoning as described in the last paragraph in that the agent’s goal is to actively alter the mental state of another individual by providing some additional information. The relevance of negation for this process lies in the assessment of what the other individual already knows (or not) in order to use communication to bring about the desired change. This pertains more to the speaker side of the interaction and this will therefore be our focus in our discussion below. There are, however, good reasons to believe that communication, at least in humans, depend on hearers’ assessment of what the speaker knows (or not) about their mental states (e.g. Tomasello ). This additional recursive layer is less important from a negation perspective and so we omit this discussion for now. On a very basic level, before thinking about how communicative acts alter the mental states of the recipient, speakers need to ensure that hearers can perceive them. Great apes

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ,  ,   . ¨ 

have been found to adjust their gestures to the attentional state of their recipient (Liebal, Call, and Tomasello ; Leavens et al. ; Call and Tomasello ). They choose more tactile gestures when the recipient is facing away and occasionally move into the recipient’s line of sight before producing visual gestures. This echoes the findings discussed above and suggests a basic understanding of the conditions that have to be met for others to have certain mental states, or not. An interesting extension of this is situations in which animals choose to communicate about certain things depending on their recipient’s knowledge state. In a recent series of studies, Bohn and colleagues found that great apes (as well as pre-linguistic infants) use pointing gestures to refer to absent entities (Bohn, Call, and Tomasello ). In these studies, participants could either point to a visible food item of rather low quality on one plate or point to another, empty, plate that previously contained high-quality food. A point to the empty plate could inform the recipient that one wants another desired food item if they know what that plate previously contained. When Bohn et al. () varied how the participant and the experimenter had interacted around the empty plate before, they found that apes would point more often to the empty plate in cases where there had been some shared interaction around its content before. In the absence of such an interaction, apes almost exclusively pointed to the visible food item. Crockford and colleagues made similar observations when studying wild chimpanzees’ alarm calling behavior (Crockford et al. ; Crockford, Wittig, and Zuberbühler ). When encountering a potential threat, chimpanzees produce alarm calls that inform individuals around them. Such alarm calling is potentially dangerous for the caller as it might attract the predator’s attention and callers should only produce them in situations in which recipients are not aware of the threat. In a series of field experiments in which they exposed chimpanzees to dummy snakes, Crockford and colleagues found that individuals would be more likely to produce alarm calls if potential recipients were unaware of the danger because they had not seen it or had been absent during previous calling events. Taken together, these studies suggest that at least some animals tailor their communicative acts to what their recipients are likely to know based on what they experienced. Both examples further suggest that the absence of a certain experience is responsible for the production of communicative acts. One way to construe the corresponding representation is to assume a negated version of the desired state that one wants to change (e.g. ‘has not seen snake’). On a proto-negation account, apes might ascribe some kind of positively defined state to their communicative partner which has the same extension (given a certain context) as the negated state described above. In the case of alarm calling this could be something like ‘thinks the path is safe’ or ‘wants to walk along the path’. These states need to be mutually exclusive with the state that the communicative act is intended to bring about (e.g. ‘thinks the path is dangerous’ or ‘wants to walk around the bush with the snake’). Again, the specific contrary pairs that would allow the use of proto-negation in these situations are discrete and formed based on an individual’s experience. This implies a very limited set of potential alternatives that could be considered and also makes the provision of fine grained and specific information less likely.

.... Summary In this section we reviewed some of the empirical evidence that relates to the question of whether nonhuman animals engage in negation proper. Many, if not all, of these findings

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



can be explained in terms of animals reasoning based on mutually exclusive states (proto-negation). These explanations assume that the animal had some specific experiences in the past that led to the formation of each specific contrary pair and/or inferred the mutual exclusivity of certain states based on, for example, some form of causal knowledge. In any case, reasoning based on proto-negation is limited to going back and forth between discrete states of affairs. As we sketched out, the main advantage animals would gain from operating on negation would be the availability of a greater number of alternatives since reasoning based on proto-negation is limited to number of available contrary pairs. The literature on inferential reasoning in nonhuman animals (for a review, see Völter and Call ) provides examples in which the performance of animals does not depend crucially on the precise stimulus. Often changing the stimuli or presenting them for the first time does not deter them from making appropriate responses. In some cases, such as the literature on inferences by exclusion, one may invoke non-inferential explanations (e.g. they are avoiding the empty cup) but some new data shows that not even this is a good explanation for their responses (e.g. Jelbert, Taylor, and Gray ). Moreover, success in other tasks such as stage  invisible displacements or exclusion tasks with two distinct hidden items (e.g. grape and banana) cannot be explained by invoking the avoidance of the empty cup. Instead, subjects’ responses appear to be governed by truly inferential processes. In section . we will build on the idea that negation, in contrast to proto-negation, might allow for more behavioral flexibility and suggest some ways of empirically approaching the question whether nonhuman animals are able to reason based on negated propositional states. We will follow the same structure as above, starting with individual reasoning and then moving on to social reasoning.

.. H    -    

.................................................................................................................................. A major challenge for future research is to find methods that would allow us to distinguish between proto- and proper negation. Here we sketch out approaches that could generate evidence that would be suggestive that an individual is negating a proposition as opposed to engaging in proto-negation. None of the approaches below would yield definitive evidence for negation proper, but with sufficient evidence accumulating, negation proper might turn out to be the better explanation for the data at hand. Most of the suggestions below refer to the idea that operating on negated propositional states (as opposed to mutual exclusive states in the world) increases the number of available behavioral responses. Furthermore, given that negation in humans is a logical operation, it is independent of the content of the propositions that are being negated. Therefore, even if this kind of evidence were found in some animal species, it would be necessary to show that the ability is not restricted to a few specific situations (e.g. reasoning about tool functionality or object permanence).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ,  ,   . ¨ 

... Individual reasoning One potential way to distinguish between proto-negation and negation proper would be to look at individual differences (see Völter et al.  for a recent plea for looking at individual differences in comparative cognitive research). As mentioned above, the respective contrary pairs that figure in proto-negation are experience dependent and therefore specific to certain situations. This implies that individuals differ in their proto-negation abilities due to the kind and amount of experience they possess, and not due to variation in a domain-general cognitive ability. Based on the proto-negation account, individuals who do well in one task should not necessarily do well in another task. On the other hand, in the case of negation proper, one applies the same logical operation to all sorts of propositions. It may therefore be regarded as a domain-general cognitive ability. This implies systematic individual differences across tasks. When testing individuals in a series of tasks involving reasoning from one state of affairs to its negated counterpart(s), the two accounts make differential predictions. Finding that animals vary systematically in their performance would at least be suggestive evidence that the same kind of logical process is involved in all cases. Importantly, consistency in performance in such tasks could originate from systematic differences between individuals in learning/inferential abilities that lead to the formation of contrary pairs. For example, Herrmann and Call () found that some great apes consistently perform better than others in a variety of cognitive tasks. These individuals might be better at spontaneously inferring mutually exclusive states and would therefore also perform better in tasks requiring reasoning from one state of affairs to its negated counterpart. Nevertheless, finding no relation between tasks would speak against a common logical basis.

... Social reasoning When discussing the literature on competitive social reasoning, we mentioned that most studies to date asked subjects to choose between one of two responses. Making the correct decision in these scenarios is well captured by the idea that animals reasoned based on contrary pairs. This conclusion could be challenged if participants were found to manipulate the experimenter’s epistemic states in multiple different ways and chose flexibly among alternatives. For example, we could imagine a scenario in which an experimenter is guarding a reward and whenever she sees the subject approaching it, she takes it away. To obtain the reward, the subject would either have to actively distract the experimenter or approach the reward in such a way that the experimenter cannot see it (see e.g. Whiten and Byrne , Coussi-Korbel , Hirata and Matsuzawa  for different examples of such tactics). By varying features of the setup, such as the presence of barriers or potentially distracting objects in the vicinity, from trial to trial, one could see if the subject adjusted flexibly to each situation. For example, crouching behind a barrier when available or throwing a ball behind the experimenter when there is one. The common thread connecting these alternatives would be that the animal is trying to keep the experimenter from entertaining a certain epistemic state (e.g.: ‘sees my hand when I reach for the food’) and alternatives might be chosen because they do not have this state as an anticipated

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



consequence. Flexible switching between alternatives would at least be suggestive that the desired effect of each action is negatively defined and therefore represented as a negated proposition instead of a positively defined state of affairs in the world. Interestingly, there is some evidence that apes flexibly switch between alternatives in the case of a positively defined state of affairs (i.e. ‘X looks at me’, see e.g. Liebal, Call, and Tomasello ; Leavens, Hostetter, Wesley, and Hopkins ). A structurally similar approach could be taken in the area of communication. In the studies mentioned above (Bohn et al. , Crockford et al. ), apes decided between producing a given communicative act (pointing to the empty plate, making an alarm call) or not. In line with a proto-negation explanation, the mental state that they ascribed to their communicative partner and which guided their decision whether or not to communicate could have been defined in a positive way as being about some concrete state of affairs in the world. From experience, they knew that a given communicative act would lead their partner to entertain the complementary (mutually exclusive) state they desired them to have. The ascription of a negatively defined, propositional state could be the more parsimonious explanation if there were evidence that animals flexibly provide information that their communicative partner lacks in situations that require multiple, complementary bits. For example, if information A and B both need to be known, they would provide A to someone who does not know A and B to someone who does not know B. It would still be possible that animals have experience with and therefore represent each of the four combinations of information A and B (AB, A¬B,¬AB, ¬A¬B) as discretely different states, it would just be less plausible, especially when more bits are added. We are aware that the approaches outlined here require additional cognitive capacities that might be beyond what animals are capable of. For example, spontaneous and flexible switching between responses requires subjects to construct these alternatives and simulate their likely effects on the experimenter. This requires a fairly elaborate theory of the experimenter’s mind. Nevertheless, we hope that these examples serve to illustrate possible ways to study how animals cope with situations that would typically involve negation in humans.

.. N      

.................................................................................................................................. We argued that the main benefit of reasoning based on negation, as opposed to protonegation, is the larger number of conclusions that can be drawn from a given premise. Proto-negation is limited to discrete contrary pairs whereas the conclusions to be drawn from negated statements are potentially limitless (in the way that the opposite of white is not black but non-white). The main prerequisite necessary for negation proper is propositionally structured thoughts. From an evolutionary perspective, emergence of negation had to be preceded, or at least accompanied, by the emergence of propositional thought (Bermúdez ). It is beyond this chapter to give a comprehensive overview of all the accounts on how propositional thought evolved. One way to think about this would be a Vygotskian view (, see e.g. Moll and Tomasello ) in which propositionality emerges first in communicative interactions as a consequence of multiple agents

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ,  ,   . ¨ 

coordinating their, sometimes differing, perspectives on one particular state of affairs. Differing perspectives encourage communicative partners to think about the specific way in which someone else sees something and creates a need to represent the corresponding intension. Such an account presupposes the emergence of fairly sophisticated socialcommunicative skills. However, once ways to represent the intensionality attached to a particular perspective in communication emerged (in vocalizations or gestures), this form of representation could subsequently be internalized and used for thought. Negation might emerge in conjunction with propositional thought because individuals need to be able to reject each other’s particular perspectives in communicative interactions. Such a rejection expresses that one thinks that a particular perspective is false as opposed to a particular state of affairs not being the case. For example, if you tell me (e.g. by pointing and pantomiming) that there is a bear hiding behind the bush but I saw the bear walk off earlier, when I reject your statement, I do not reject that there is a bush but your perspective that there is a bear behind the bush. The emergence of negation and propositional thought would be intimately linked in this case.

.. C

.................................................................................................................................. To date, there is no compelling evidence for negation proper in nonhuman animals. This conclusion is based on a particular perspective on negation, namely that it involves the negation of a particular proposition and is therefore tied to propositional thought more generally. As we have seen, there is plenty of evidence suggesting that animals engage in forms of reasoning that are functionally related to negation but might be procedurally different. Nonhuman animals might engage merely in proto-negation (Bermúdez ), that is reasoning based on mutually exclusive contrary pairs. However, behavioral evidence is often ambiguous with regard to the underlying cognitive processes and there is accumulating evidence for high degrees of behavioral flexibility in animals also in situations involving negative information (e.g. about the absence of a stimulus) and so the jury is still out. Joint efforts from comparative psychologists, philosophers, and computational cognitive scientists will allow us to specify how and when negation (and propositional thought) emerged in phylogeny.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ......................................................................................................................

                       ......................................................................................................................

-´  

. I

.................................................................................................................................. D decades of research investigating statistical learning (Smith and Yu ), hypothesis testing (Trueswell et al. ), and fast mapping (Carey and Barlett ; Swingley ), much remains to be understood about the mechanisms that underlie word learning. Moreover, much research has focused on how the name for object categories (i.e. nouns), observable actions (i.e. action verbs), or perceptual properties (e.g. color adjectives) are acquired. Little is known of how the meaning of more abstract words is acquired, with the notable exception of integer words (Carey ). The mystery is particularly thick when it comes to function words (Hochmann, Endress, and Mehler ; Hochmann ), among which are the words that express negation (e.g. ‘not’). Those words do not refer to observable objects, actions, or properties, not even to abstractions from such observable entities. Rather, they express a computation on the meaning of associated words. Negation, in particular, flips the truth values of the proposition it combines with. In this chapter, we will not attempt to solve this mystery. In this respect, we will only argue that the meaning of a word is not magically formed when the corresponding phonological form is acquired. Rather, for any meaning that children or adults acquire, there must exist at least one pre-lexical precursor representation. In some cases, the precursor mental representation may be extremely close to the meaning eventually expressed by the word (e.g. possibly the precursor to the meaning of the word ‘circle’ is a mental representation of a circular shape). In this case, learning the meaning of a word consists in mapping a word form to an existing mental representation. In other cases, the precursor mental representation may consist in disconnected features that need to be combined or bounded together (e.g. possibly the meaning of the word ‘heater’ is composed of the idea of an object, the idea of causality, and the idea of heat). In this case, learning the meaning of a word involves a representational change, as a new mental representation is formed by combining, binding, or modifying existing mental representations. Our goal in

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



-´ 

this chapter is to investigate what precursors of negation may exist in the mind of young infants, before they learn the meaning of a word like ‘not’. This exploration, of course, must be guided by the meaning of the words for negation (see also Chapter  by Prieto and Espinal on the role of gestures in expressing negation). In section ., we will briefly discuss the meanings that young children initially assign to words such as ‘no’ or ‘not’. In sections .–., we will review the extant infant cognition literature, searching for precursors of these meanings. Finally, in section ., we will discuss the role that inhibition may play in the acquisition of verbal negation.

. E   

.................................................................................................................................. Three broad categories are generally accepted to classify the meanings of negation when it first appears in children’s verbal productions: () rejection or refusal, () disappearance, non-existence, or more generally unfulfilled expectation, () truth-functional negation or denial (Dimroth ; Pea a; Choi ). These three categories vary in the complexity of mental representations that they require. Rejection () does not require any lasting mental representation, but solely an attitude toward a contextual external object. One situation familiar to all parents exemplifies this category: a parent approaches a spoon filled with homemade apple sauce, in order to feed her child. The child vehemently shakes her head, closes her lips, or verbally expresses ‘no!’ In this situation, the child is refusing/rejecting the food that is presented to her. But that behavior is only directed towards an external object, as long as it is present. As soon as the spoon is taken away, the rejection behavior ceases. Unfulfilled expectations (), especially disappearance, in contrast, require a mental representation that is dissociated from contextual reality, but that representation is a positive one. Here, a verbal utterance such as ‘no’ expresses the detection of a mismatch between a positive expectation and the observed state of the world. For instance, imagine a situation where a child expects to find a ball in a box. Her mental representation is that of a ball in the box. Uttering ‘no’ when opening the box to find it empty expresses the mismatch between her expectation and reality. Finally, truth-functional negation or denial (), which we assume to be the adult meaning of the word ‘not’, relies on the relation of two opposite mental representations: what is and what is not. It is observed when an adult expresses a proposition; for example “this ball is blue,” and the child expressing her disagreement: “no!,” especially if she adds “it’s green.” In this situation, the child represents a relation between the actual color (green) of an object and another color (blue). In contrast to the previous category, where the mental representation was positive and contrasted with reality, here the mental representation is negative. Children represent the color of the ball in its negative relation to the color blue. That type of negation is last to emerge in the language of young children (Dimroth ). Of course, distinguishing these three types of negative meanings is not always easy. Moreover, it remains possible that a single truth-functional meaning could underlie all of toddlers’ and young children’s utterances (Feiman et al. ). To better identify the development of truth-functional negation, a recent study tested children’s understanding

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



of negative utterances in the context of logical reasoning (Feiman et al. ; see also Austin et al. ). In that experimental manipulation, a ball was hidden in one of two containers, a truck or a bucket. Children received either a positive cue or a negative cue as to the position of the ball (“it’s in the truck”; “it’s not in the bucket”). Results show that children are unable to use the negative cue to exclude one of the two alternative positions and infer the actual position of the ball, before the third year of life ( months). An aspect that could contribute to these results is that younger children may consider it pragmatically strange to give a cue in a negative rather than a positive form. Taking this into account, Feiman and colleagues replicated their results presenting cues as the answer to a question, which is better pragmatically (“Is it in the truck?” “Yes, it is” / “No, it’s not”). Again, -month-olds, but not younger children, searched correctly. The task could potentially be solved with truth-functional negation or with a nonexistence meaning of negative words. However, the comprehension of negative utterances in this task emerges at the same age at which children begin to produce denial/truthfunctional negation, whereas non-existence meanings are attested earlier. Altogether, these results strongly suggest that truth-functional verbal negation expresses a meaning that is qualitatively different from the other two categories (rejection and non-existence/unfulfilled expectation), and does not emerge before the third year of life. In sections .–., we will review the literature on infant (<  months) cognition, searching for evidence for precursors of each type of early meaning of verbal negation: rejection (section .), unfulfilled expectations (section .) and truth-functional negations (section .).

.. A     

.................................................................................................................................. Avoidance responses surely exist independently from language. Such responses have been observed in conditioning experiments, where an animal is trained to avoid a certain stimulus or behavior, in order to avoid negative reinforcement in the form of an electric shock for instance (Solomon and Wynne ). Even spontaneously, in the absence of conditioning, all animals tend to avoid certain sensations, especially certain smells. That kind of behavior is often taken advantage of to keep pests or insects away. Similarly, infants perform spontaneous avoidance behavior in certain situations. In the famous visual cliff experiments (Gibson and Walk ), for instance, -month-old and older infants tend to avoid walking over a high drop-off (the cliff) if they perceive it is too high for their body and posture (Gibson ; Adolph, Kretch, and LoBue ). Recently, Hochmann, Mody, and Carey () found that -month-old infants could learn to avoid a particular behavioral response. In a non-match-to-sample task (NMTS), infants were presented with a series of animated movies that always followed the same structure (Figure .). Three cards were presented on the screen, each displaying a symbol. Two of the cards displayed the same symbol (either the central card and the left card, or the central card and the right card), and the third card was different (Figure .A). After each card was displayed, the different card was animated. The final card to be displayed was



-´  A

B

Training

Test visible-Same

Test visible-Different

Correct anticipations

Infants correctly avoid the “same” card

Random anticipations

Time

 .. Structure of the training (A) and test (B) trials of the Non-Match to Sample task (Hochmann, Mody, and Carey ).

always the central one, which was never the different one. Thus, infants could not predict which card would be animated before the central card was displayed. Data shows that infants learned the structure of these movies. They correctly anticipated the final animation, looking at the different card before it was animated. At first glance, these results suggest that infants learned to choose the different card. However, additional test trials show that infants learned an alternative rule. Infants learned to avoid the same card. Indeed, in the test phase (Figure .B), when either the identity of the left or right card was not displayed, infants continued to anticipate correctly the animation, but only if they saw the card that was the same as the central card. In this situation, they anticipated the animation of the other card, the one that was never shown. In contrast, when infants saw the different card, they searched randomly, suggesting that they did not know that card should be the animated one. In sum, these results suggest that infants learned a rule that is based on avoidance, rather than a rule based on a positive choice. Interestingly, whereas infants are able to learn an avoidance rule that applies to themselves, they seem unable to interpret the behavior of others as guided by avoidance. Feiman and colleagues () tried to teach - and -month-old infants that a third agent confronted with repeated choices between two objects would always avoid one specific object. Infants appeared unable to make sense of that behavior. They could, in contrast, easily make sense of a positive behavior, for example systematically choosing one specific object. In sum, avoidance appears to be available to guide infants’ behavior but not to make sense of others’ behavior.

   



.. U / 

.................................................................................................................................. Many experiments on infant cognition rely on a paradigm called violation of expectation, in which researchers analyze infants’ reactions to impossible or unlikely events in order to infer their expectations (e.g. Baillargeon, Spelke, and Wasserman ). Increased looking times to certain events are interpreted as a sign of surprise, suggesting that the presented event has violated infants’ expectations. This response thus suggests that infants have the capacity to detect a mismatch between their expectations and the current state of the world. For instance, in Baillargeon’s seminal experiment, an object is placed behind a rotating screen. In some events, the rotating screen was raised up until it covered fully the view of the object for the infant, continued rotating, and stopped when it contacted the hidden object. In other events, instead, the screen continued its rotation, as if the hidden object had been removed, flattened, or ceased to exist. Five-month-old infants looked longer at the second type of events, suggesting they expected the screen to stop when contacting the hidden object, and revealing that they knew the object continued to exist even though it was no longer visible for them. Of utter interest here, are experiments on object representations, where a number of objects are placed behind an occluder. When the occluder is removed, infants look longer if the number of objects has varied. For instance, -month-old infants look longer if they expected two objects and find only one or three (Wynn ). Eight-month-olds also looked longer when they expected one object and found none behind the occluder (Wynn and Chiang ). This situation is analogous to the situation that classically elicits non-existence utterances in young children. While infants can react to the absence of an object, an interesting failure suggests that they cannot expect the absence of an object. Indeed, when infants should expect no object at all behind an occluder (i.e. because they saw the object had been removed) and find one, no increased looking time was observed (Wynn and Chiang ; see also Kaufman, Csibra, and Johnson ). Of course, a non-result is never conclusive. Still, these experiments suggest that it is difficult for infants to represent the absence of an object. They can contrast perceived absence with an expected presence, but not perceived presence with an expected absence.

.. A    

.................................................................................................................................. Truth-functional negation, as one would expect, is more controversial in preverbal infants. A first strategy, to uncover a precursor of truth-functional negation is to investigate the inferential role it may play in logical reasoning tasks. Toddlers as young as  months and a number of non-human species (e.g. Call ; Sabatini and Visalberghi ) are able to



-´ 

solve a non-verbal disjunctive syllogism akin to the verbal version discussed in section . (Feiman et al. ). Toddlers witness a toy being hidden in one of two cups, but because the action is partly occluded, they initially ignore in which one. When experimenters show that one cup is empty, toddlers correctly search for the toy in the other cup (Mody, Feiman, and Carey submitted), suggesting that they expect to find the toy there. This behavior could be interpreted as the trace of the inferential role that truth-functional negation can play (“if the toy is not in the left cup, it must be in the right cup”). However, alternative accounts must be considered. In a recent paper, Mody and Carey () argued that toddlers may in fact use a simpler strategy: avoiding searching in an empty cup. In other words, toddlers may search in the correct cup, not because they know the toy is there (which would be the correct logical inference), but because they have no other place left to search. They thus designed a version of the task, where solely avoiding the empty cup would not necessarily lead to the correct choice. The clever idea consisted in doubling the number of toys and cups. One toy was hidden in one of two cups (cup A or cup B) and a second toy was hidden in one of two other cups (cup C or cup D). When one cup was revealed to be empty (e.g. cup A), avoiding that cup would lead to choosing randomly between the three other cups (B, C, and D). But reasoning logically, possibly using truth-functional negation, would lead to the conclusion that if cup A is empty, cup B necessarily contains a toy. Three-year-olds, but not younger children, succeed at this task, searching correctly in B more often than in C or D. The age difference in success observed for the two- and four-cup tasks may be due to an increase of complexity and working memory demands, but it may also reveal something about the development of truthfunctional negation. In fact, success at the four-cup task appears to come later than the production of truth-functional negation (in the third year of life). Recently, Cesana-Arlotti and colleagues () asked whether - and -month-olds could solve another disjunctive syllogism task, which avoids the pitfall of the “avoiding the empty cup” strategy. In their clever paradigm, two objects (e.g. an umbrella and a hat) were initially presented on the screen. These objects were only distinguishable by their lower halves but shared identical top halves. The two objects were then occluded by a screen, and one object was placed in a cup, so that only the top half was visible. At this point, infants ignored which object was behind the screen and which object was in the cup. When the identity of the object that remained behind the screen was revealed, an inference was possible. If the hat remained behind the screen, then the umbrella was not behind the screen and must be in the cup. Findings suggest that infants reached the correct conclusion. This task thus follows the same abstract structure as the two-cup task and the situation used by Feiman and colleagues in their study of verbal negation understanding. Two possibilities are initially possible, A or B (the umbrella or the hat; one cup or the other; in the bucket or in the truck). The exclusion of one possibility leads to the correct inference. In the verbal version of the task (Feiman et al. ), success appears to require the truth-functional meaning of verbal negation, as children do not succeed before they produce such verbal negation. By analogy, does success in the non-verbal version of the task (Cesana-Arlotti et al. ) signal a precursor of truth-functional negation? These results are certainly suggestive of a precursor of negation that could be called exclusion, but additional evidence is necessary before we can definitely link it to truth-functional negation.

   



A second strategy, proposed by Bermúdez () in the study of animal cognition, is to begin by trying to find evidence for infants’ representations of pairs of concepts linked by negation: for instance presence/absence, same/different. The strategy would be to first find evidence for the representation of two contradictory concepts, and second to ask whether one is indeed represented as the negation of the other. As detailed in section ., it appears that infants represent presence, but have trouble representing absence, raising doubts on the availability of truth-functional negation. Recent studies on the pair of relational concepts same and different have addressed this issue in more depth. Results convincingly show that infants have a representation of the relation same. For instance, infants can learn an arbitrary rule, such that the stimulus that is the same as an exemplar is going to be animated, generalizing the rule to novel stimuli (Match-to-Sample task; Hochmann, Mody, and Carey ). Infants can also learn that the presentation of two stimuli that are the same (e.g. two same syllables: ba ba; two same Types of syllable sequences A

A

A

A

A

A

A

B

A

A

B

B

25% 25% 25% A 0.3

A

A

cluster mass test P < .01

A

25% *

variation of pupil diameter (mm)

0.2 0.1 0 –0.1 –0.2 –0.3 –0.4 –0.5 –0.6 –2500 –2000 –1500 –1000 –500

0

500

1000 1500 2000

2500

3000 3500 (ms)

 .. Results of an oddball experiment relying on pupillometry (Hochmann and Toro submitted). The curves represent the average variations of pupil diameters (mm) with respect to a  ms baseline taken before the onset of the fourth syllable. The grey area depicts the time windows where a cluster mass permutation test detected significantly (*P < .) larger pupil dilation for deviant trials (red) with respect to standard trials (blue). Light-colored areas represent standard errors from the mean. The crenel functions indicate when syllables occurred within each trial.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



-´ 

vowels: ba da; two same geometrical shapes: ) predicts the location of a puppet’s appearance (Hochmann et al. , ; Hochmann, Carey, and Mehler ; Kovács ). These and all the relevant results in the infant literature (Ferry, Hespos, and Gentner ; Walker and Gopnik ) can be accounted for by a representation of same relying on representations of variables for individuals: for X ∈ Σ, (X X), where X is a variable for an element of Σ, the ensemble of stimuli relevant to the task (e.g. all syllables; all visual shapes, etc.) In contrast, the evidence is sparse for the relation different. As discussed in section ., infants can learn to look at the stimulus that is different from an exemplar (Non-Match-toSample task), but further tests show that the mental rule infants represent is predicated on the relation same, not on the relation different (Hochmann, Mody, and Carey ). Indeed, infants learn to avoid the stimulus that is the same, rather than choosing the stimulus that is different (see also Zentall et al. , ). Furthermore, infants fail to learn a rule such that the presentation of two different stimuli predicts the location of a puppet’s appearance, but this may be due to the difficulty of learning in parallel the rule about same and the rule about different (Hochmann, Carey, and Mehler ). At this point, there is no convincing evidence, that infants have a representation of the relation different. Nevertheless, new results from our lab suggest that infants are able to represent the relation different after all (Hochmann and Toro submitted). We used an oddball paradigm, where stimuli consist in sequences of syllables. In our paradigm, most sequences followed the same structure: a sequence of same syllables, ending with a different syllable (e.g. ba ba ba di; ko ko se; lu lu lu lu na; etc.). Once in a while, however, we presented a sequence that violated the common structure, removing the final different syllable. Infants exhibited surprise in response to these sequences, as signaled by pupil dilation (Figure .). The results of this experiment suggest that infants learned to expect a final different syllable. These results suggest that infants do possess a representation of the relation different. Now, is showing that infants represent both the relation same and the relation different sufficient to argue they have a precursor of truth-functional negation? We would like to argue that it is the case. As discussed in section ., the key aspect of truth-functional negation is that it is relational. The representation of the relation different that we have uncovered exhibits this aspect. Indeed, when infants learned that the sequence of syllables starting with mi mi mi should finish with a syllable that is different from mi, it means that they are ready to accept any syllable but the syllable mi. In other words, infants have no idea what the final syllable will be, except that it cannot be the syllable mi; they define the final syllable in a negative relation to the syllable mi. In this sense, we believe that the infant representation of the relation different suggests the existence of a precursor of truthfunctional negation.

.. I     

.................................................................................................................................. In the previous sections, we exposed the evidence for precursors of the successive meanings of verbal negation in preverbal infants. We argued that there exist precursors for avoidance, non-existence, or unfulfilled expectations, and even truth-functional negation.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



How could truth-functional negation be implemented in infants’ minds? Our experiment presented in detail in section . suggests that infants are capable of expecting that a syllable could be anything but a specific syllable (e.g. anything but mi). Computationally, this could be implemented by inhibiting the representation of the syllable mi for the mechanism responsible for predictions. Thereby, all syllables remain possible and predictable, except the syllable mi. These considerations led us to ponder the role that inhibition could play in the development of negation in infancy. In this section, we would like to propose an untested scenario that would gradually transform an inhibitory response into truth-functional negation. By the end of the first year of life, an infant comes to understand the exclamation “no!” When hearing that exclamation, she interrupts her current activity and looks at the person (often a parent) who produced the exclamation for more information. Her mom may point at the plug the infant was approaching and repeat “no!” She may also give linguistic information “Do not put your fingers into the plug.” My colleague Liuba Papeo and I propose that this type of interaction may participate in the development of negation in three steps (see Russell  for a similar untested proposal). ) An innate alarm call, selected through evolution, could elicit a freezing response in infants. Such call and behavior are common in animal species (e.g. Langbauer et al.  for elephants; Mateo  for squirrels; Smith  for chickadees). In humans, the specification of such a freezing call may correspond to the universal intonation in the exclamation “no!” (Fernald ). Physiologically and cognitively, the freezing response corresponds to a general inhibition of the current action and mental representations. ) Infants may learn to associate the freezing response to the lexical entry ‘no’, even in the absence of the prosodic properties of the exclamation “no!” That is, the freezing response is now elicited by the word ‘no’. ) Gradually during development, using the additional information provided by the person who first produced “no!,” infants may learn to constrain the inhibition to a particular behavior or representation, as opposed to a general freezing response. For instance, in the situation described above, the infant may learn that inhibition should be applied to the specific act of approaching plugs. In this process, the general inhibitory response to alarm calls becomes an operator that acts on an argument. This hypothesis is so far untested, but it makes predictions. In particular, if inhibition plays a role in the meaning of verbal negation, this should still be observable in adults. In consequence, we should observe activations of inhibitory mechanisms when processing negative sentences. Recent results investigating the neural mechanisms of negation lend credit to the hypothesis (see Chapter  in this volume; Papeo, Hochmann, and Batelli ), calling for further investigations.

.. C

.................................................................................................................................. Negation, an operator that flips the truth value of a proposition, cannot be created ex-nihilo when words such as ‘no’ or ‘not’ are acquired. Rather, there must exist pre-lexical

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



-´ 

precursors of negation in infants’ minds. In this chapter, we have reviewed the infant cognition literature in search of such precursors. We showed that there exist precursors of early meanings of verbal negation such as avoidance and the detection of mismatches between expectations and reality. We further argued that the infant representation of the relation different suggests the existence of a precursor of truth-functional negation. Finally, we proposed a testable hypothesis on the role of inhibition in the development of verbal truth-functional negation. This brief review of the literature, we hope, can set the ground for further investigations on the cognitive implementation of negation and its development in the first years of life.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ......................................................................................................................

                     ......................................................................................................................

 

.. F    ’   

.................................................................................................................................. E work on children’s acquisition of negation focused on how children developed the mapping between negative expressions and their different semantic/pragmatic functions. Researchers studied children’s correspondence conditions between negative expressions and identified as many as eight semantic/pragmatic functions (Choi ; CameronFaulkner, Lieven, and Theakston ). Three functions featured prominently in early research: children’s use of negation to express refusal, non-existence, and denial. A study of three English-speaking children by Bloom () concluded that each child progressed through the same three stages. At the first stage, children used negation to express nonexistence. This was followed by a second stage at which negation expressed refusal and a third stage expressing denial. The same order of acquisition was found for children acquiring French and Korean (Choi ), and for children acquiring Cantonese (Tam and Stokes ). The focus on the function of children’s negative sentences meant that some of the early research did not differentiate children’s use of negation as anaphoric negation or sentential negation (e.g. Cameron-Faulkner, Lieven, and Theakston ). Anaphoric negation is generally produced in response to a preceding utterance. In English, it is typically expressed by the negative marker no, as illustrated in (). ()

Parent: Did you eat your beans? Child: No! . . . I didn’t eat my beans.

Sentential negation targets sentence internal propositions. It is typically expressed by not and by auxiliary verbs accompanied by the cliticized form of negation, n’t. The continuation . . . I didn’t eat my beans in example () illustrates sentential negation, where the negative auxiliary verb didn’t negates the proposition I ate my beans. The distinction between anaphoric and sentential negation is somewhat murky in the literature on children acquiring English. This

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

is, in part, because some English-speaking children initially use no to express both kinds of negation, and in part because children’s early productions consist of just two or three words, thereby obscuring the position of negation in children’s sentence structures. Nevertheless, Bloom () claimed that context notes could be used to distinguish both anaphoric and sentential negation in young children’s grammars, as illustrated in (), where (a) is presumably an instance of sentential negation and (b) is anaphoric negation. ()

a. Kathryn no fix this Context: Kathryn could not fix the train b. No Kathryn playing self Context: Kathryn did not want to play with the slide and went into the playpen to play alone (Kathryn, Time ,  months)

In what follows, we draw attention to children’s acquisition of various aspects of sentential negation, including cross-linguistic research on the role negation plays in children’s developing grammars.

.. E    

.................................................................................................................................. Children acquiring Romance languages are reported to use negation early in the course of acquisition. For example, Grinstead () notes that the three Spanish-speaking and four Catalan-speaking children in his data sample use sentential negation very early in declarative sentences, as illustrated in () and (). ()

()

Spanish No tiene miedo Not has (sg. Pres) fear ‘He isn’t afraid.’ Catalan No vol. Not want (sg. Pres, declarative) ‘He doesn’t want to.’

(Graciela ;.)

(Pep ;.)

Italian-speaking children also produce negation early. An investigation of verbal syntax in Italian-speaking children by Guasti () established early mastery of adult-like word order in negative sentences. Guasti argued that children’s use of the word order non+V +negative adverb in their productions is compelling evidence that they know the hierarchical position of negation in sentence structures. In adult Italian, the negative clitic non initially raises from Neg to I, and then cliticizes to the left of the verb which has itself raised from V to I. By contrast, negative adverbs such as più ‘anymore’ and mai ‘never’ are

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

    



positioned in SpecNegP, and thus follow the verb. Children’s use of multi-word utterances with the order non+V+negative adverb suggests that both the verb and the negative clitic non have raised to I in the hierarchical structures that children assign. Some early examples are given in () and (). Example () has a null subject and a missing negative clitic, although the negative adverb is present. The negative clitic is present in example (), alongside the negative adverb. ()

()

_ (v)oio più _want-sg anymore ‘(I) do (not) want anymore.’

(Martina ;)

ora no piange più Now Neg cry-sg anymore ‘Now (she) does not cry anymore.’

(Martina ;)

It has proven more difficult to determine when English-speaking children position negation in the same way as adults, partly because verb movement cannot be used as a diagnostic for the hierarchical position of negation (although children do not produce utterances with a main verb to the left of negation, so we can surmise that verb movement has not taken place). The lack of a definitive diagnostic may, in part, explain why there is disagreement in the literature on the position of negation in the early grammars of Englishspeaking children.

.. E   

.................................................................................................................................. In seminal work, Bellugi () and Klima and Bellugi () concluded that Englishspeaking children’s development of negation consisted of three stages. This conclusion was based on transcripts of the Harvard children—Adam, Eve, and Sarah (Brown ). At Stage , sentential negation was expressed by no or not, and appeared external to the remainder of the child’s utterance, either preceding or following it, as illustrated in () (Bellugi ). At the second stage, negation appeared mostly sentence internally, as illustrated in () (Bellugi ). Children reached stage  once they attained productive use of a range of negative auxiliary verbs. ()

a. b. c. d.

No the sun shining Take it . . . no No pants No Mommy read

(Adam) (Adam) (Sarah) (Eve)

()

a. b. c. d.

That no blast off He don’t want some money He no bite you Not go in there

(Adam) (Adam) (Sarah) (Eve)

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

Adopting a transformational approach to child language, Klima and Bellugi () proposed that, at Stage , children position no or not outside of a sentence “nucleus.” According to Bellugi () Stage  negation is “primitive” and has “little internal structure,” consisting primarily of unmarked nouns and verbs (Bellugi : ). More complex examples such as No the sun shining (a) have led to a variety of hypotheses both about sentence structure and its pragmatic function. Because negation appears outside the sentence “nucleus,” it is unclear where it resides structurally in these utterances. Some researchers have suggested that negation can reside outside the sentence itself. See also Gilkerson, Hyams, and Curtiss () for this view. Bloom () argued against this position, on grounds of Continuity. Bloom argued that it is more parsimonious to view the child’s grammar as qualitatively similar to that of adults. A different approach was adopted by Drozd () who argued that examples such as (a) express “metalinguistic exclamatory sentence negation” and not sentential negation. For Drozd (), negation served a pragmatic function found in colloquial adult conversations (cf. Horn ). For Drozd, the continuity of syntax is exchanged for continuity of function. Recent evidence has surfaced that supports the proposal that children’s use of external negation can apply to propositions. If so, the examples in () express sentential negation, as Bellugi () argued. The evidence for external negation comes from the productions of a child who uses home sign (Franklin, Giannakidou, and Goldin-Meadow ), a gestural system claimed to share many of the characteristics of natural languages (Goldin-Meadow ). This child was recorded at eight points in time, between the ages of ; and ;. During this period, the child produced  instances of negation, serving the pragmatic functions of rejection, non-existence, and denial. Of these,  were multi-gestural, providing useful information about the position of negation in the sentence structure. The child conveyed a negative meaning by shaking his head from side to side. The side-toside headshake consistently preceded or followed sentences, and was never introduced in the middle of an utterance. In the child’s multi-gesture productions, % of side-to-side headshakes expressing negation were introduced before the sentence, with the remaining % at the end. Importantly, when negation was introduced at the beginning, it frequently continued throughout the utterance. This suggests that negation applied to propositions expressed within the child’s sentences. The authors take this appearance of external negation as evidence of one typological possibility for marking negation across languages.

.. N, ,  

.................................................................................................................................. Although it seems clear that sentence external negation is possible in natural languages, and may well appear in early child language, the optimal analysis of sentences like No the sun shining remains controversial. An account by Déprez and Pierce () denies that negation is external to the sentence in such utterances. Drawing on the VP-internal subject hypothesis, these researchers contend that productions like No the sun shining arise because children fail to raise the Subject NP out of its VP-internal base position to the subject position in SpecIP. The VP-internal position of the Subject NP explains why negation appears to the left of the subject in examples like (a). The Déprez and Pierce analysis is schematically depicted in (), where it can be seen that the Subject NP the sun is positioned

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

    



in SpecVP. One advantage of this analysis is that it adheres to the Continuity assumption since child and adult grammars are characterized by the same sentence structure. ()

[IP [NegP no [Neg [VP the sun [V shining ]]]]]

Assuming that the Subject NP is often omitted in young children’s grammars, moreover, the Déprez and Pierce analysis explains why young children produce utterances such as No singing song.¹ As Déprez and Pierce () observe, since French is a verb raising language, the main verb precedes the negative marker pas in negative sentences; finite main verbs raise from V to I, as in (a), higher than negation. The Subject NP is null in (a), but it can surface in VP-internal position, as shown in example (b). On the other hand, infinitives fail to raise, so infinitives surface after pas, as illustrated in (c). In both cases, the Subject NP follows negation if the Subject NP remains in VP-internal position. ()

a. veux pas lolo want not milk ‘I don’t want milk’

(Nathalie ;.)

b. fait du bruit la voiture makes noise the car ‘The car makes noise’

(Philippe ;.)

c. pas manger la poupée not eat the doll

(Nathalie ;.)

According to Déprez and Pierce (), German-speaking children also initially position negation correctly. Assuming that the Subject NP may reside in VP-internal position, a VP-internal Subject NP is expected to follow negation; otherwise, the Subject NP is expected to precede negation. Verb placement in German-speaking children’s grammars depends on whether or not they have mastered V-movement. Déprez and Pierce () propose that V is delayed in some children, which results in the verb moving only as high as I in the sentence structure. It follows that the verb will appear in final position, given that the IP projection is head final. According to Hamann (: , fn.), however, children’s productions with a final finite verb are infrequent. Example () is offered by Hamann () as a sentence in which a finite verb appears in second position (V) in the speech of a young German-speaking child. ()

ich mach das nich I do that not ‘I don’t/won’t do that’

(Andreas ;)

Such examples, Hamann argues, show that verb movement is permitted across negation. This fact is used to argue that nicht cannot be a head, but must be in a specifier position, or ¹ In sentences with a null Subject NP, Infl governs and licenses the pro subject, which is argued not to require nominative Case (Déprez and Pierce : ).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

adjoined as an adverb (Hamann : ). As an adverb, negation could be adjoined in the structure to VP. Hamann speculates that it could also be, then, that in German (and English too), the adverbial form of negation can be adjoined as a sentential modifier, which would explain productions noted in the literature of children using nein or no in sentenceinitial position. Thus Hamann supports the proposal that negation can be external to the sentence in early grammars. To account for sentences with apparent ‘misplaced’ negation in young Korean-speaking children, Hagstrom () advances a proposal that is similar in spirit to Déprez and Pierce () but differs in detail. Hagstrom, too, argues that negation only appears to be misplaced in children’s sentences. Appearances are deceiving because, on his account, children retain lexical material in the VP rather than repositioning it as in the adult grammar. According to the literature, Korean-speaking children allow negation to move away from the verb in structures using the ‘Short Form of Negation’. (There is also a ‘Long Form of Negation’ that appears much later in children’s productions, but this will not be discussed here.) The adult version of the Short Form of Negation is shown in (); here negation appears immediately preceding the verb (Cho and Hong ; Kim ). The corresponding child utterances are shown in () (from Cho and Hong , cited in Hagstrom: : –). The negative marker an and the verb are separated by an Object NP in (a), and by an adverbial element in (b). () Chelswu-ka ppang-ul an mek-ess-ta Chelswu-nom bread-acc neg eat-past-decl ‘Chelswu didn’t eat the bread.’ () a. na an pap mek-e I neg rice eat-decl ‘I do not eat rice’ b. na an cal hay I neg well do ‘I do not do well’

(Child ;.)

(Child ;.)

Hagstrom () argues that the apparent misplacement of negation is due to the fact that children have failed to move lexical material, in this case the object or the adverb, out of the VP, as is obligatory in the adult grammar. This movement would leave negation correctly positioned next to the verb. Like Déprez and Pierce (), his conclusion is that negation is correctly positioned in the hierarchical structure.

.. C     

.................................................................................................................................. As compared to children acquiring Romance and other Germanic languages, children acquiring English take a considerable time to converge on the adult grammar of negation. If Déprez and Pierce () are correct, however, this cannot be because English-speaking

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

    



children initially misplace negation. This leads us to ask why the acquisition process is prolonged. Thornton and Tesan () argue that the delay has its source in children’s difficulty identifying that English has a negative head in addition to the adverbial form of negation. These researchers invoke a theoretical learnability proposal by Zeijlstra (a, b) to explain the difficulties experienced by English-speaking children. Following Zeijlstra (b), Thornton and Tesan () assume that for reasons of computational efficiency, children do not initially posit the full range of functional categories. The proposal is that the functional categories required by a particular language are triggered by positive evidence. By default, no matter what language they are acquiring, all children should initially take negation to be an adverb because adverbs can be positioned in the hierarchical structure without having to posit a NegP functional projection. The default choice is correct for children acquiring Dutch or German and no further adjustments need to be made. Children acquiring languages such as French, or Italian, where negation is a head, will quickly be exposed to the positive input they need to (re)categorize negation as a head and add a functional category for negation, NegP. On this account, the proposed positive input for NegP is negative concord. The situation for English-speaking children is more complicated, since English has both adverbial and head forms of negation (Haegeman ; Zanuttini ; Zeijlstra a). The learnability scenario for Englishspeaking children proceeds as follows. Like all children, English-speaking children initially hypothesize that English has an adverbial form of negation. According to Thornton and Tesan () and Thornton and Rombough (), most children use the negative adverb not, and some children use no as their adverb.² There is a striking consequence of having only an adverbial form of negation. Children are forced to produce non-adult sentence structures, such as (a,b) because main verbs (other than have and be) must be negated using an auxiliary verb. Although at this stage of development (Klima and Bellugi’s Stage ), children sometimes produce can’t and don’t, Thornton and Rombough () offer several reasons for analyzing these lexical items as fixed adverbial forms of negation, as Bellugi () did, rather than as true auxiliary verbs that are decomposed into Aux+n’t (but see Harris and Wexler  and Schütze  for a different view). ()

a. He no bite you b. Pink Panther not fit

To achieve the adult grammar for negation, English-speaking children must incorporate the head form of negation, n’t, into their grammar. For children acquiring Standard English, the evidence cannot be negative concord. Partly for this reason, perhaps, English-speaking children take a considerable period of time before they identify the morpheme n’t as a head form of negation. This may seem counter-intuitive, because negative auxiliary verbs should provide abundant relevant input. As we have seen, however, children initially fail to decompose negative auxiliary verbs like don’t and can’t into Aux + n’t. Children’s transition to a grammar with a head form of negation appears to await the

² Whether not is a head or an adverb is subject to debate. For present purposes, we will simply follow Haegeman (), Zanuttini (), and Zeijlstra (a) in analyzing not as an adverb, and n’t as a head.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

realization that complex negative auxiliary verbs like doesn’t are decomposed into three morphemes, Aux + -s + n’t (Thornton and Tesan ). See also Yang () for further evidence for the lack of productivity of n’t as an affix. However, figuring out the syntactic category of the negative marker may not be the only difficulty that children encounter in mastering negation.

.. F   

.................................................................................................................................. Children are less able to accommodate to infelicity than adults (Crain and Thornton ). One consequence is that when a negative sentence is produced ‘out of the blue’, young children struggle to respond in an adult-like manner. For example, Nordmeyer and Frank () cite an experiment by Kim () in which the experimenter points to an apple and says “This is not a banana.” In this situation, most of the -year-olds and half the -yearolds said this was wrong. The stimuli were, of course, presented in an unnatural context. As has been pointed out, negative sentences are only felicitous in contexts in which the corresponding affirmative proposition is under consideration (e.g. Wason ). In Kim’s experiment, there was no discussion of whether or not other objects in the context were bananas so the negative sentence was infelicitous. In order to satisfy the felicity conditions for negation in a production experiment eliciting negative sentences, Thornton and Rombough () always introduced the corresponding affirmative sentence first. As shown in the sample protocol in (), in which children were investigating which toys fitted through the door of a toy bus, children were always directed to test and comment on toys that did fit through the door before they tried ones that did not. ()

Experimenter: So, can you tell me if Minnie Mouse fits through the door? Child: She fits. Experimenter: OK, now try Cookie Monster. See if he fits. Child: Cookie Monster fits. Experimenter: OK, how about Big Bird? Child: No! Experimenter: So, let’s see, Minnie Mouse fits, and Cookie Monster fits too. What about Big Bird? Child: He not fit [ . . . ]

The Thornton and Rombough experiment elicited negative sentences from twenty-five - to -year-old English-speaking children (mean age ;). In this felicitous context, all of the children were able to produce negative sentences, although their form was not always adult-like. Thirteen of the twenty-five children produced adult-like sentences that contained the negative auxiliary doesn’t, while the other twelve children produced virtually no such sentences. Instead, these children produced a variety of non-adult sentence structures with what could be taken to be adverbial negation, as in (). These options disappeared once children acquired doesn’t.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

     ()



a. He not/don’t fit b. He’s not/don’t fit c. He not/don’t fits

There are a number of recent studies investigating children’s awareness of the pragmatics associated with use of negation using comprehension techniques (e.g. Nordmeyer and Frank , ; Reuter, Feiman, and Snedeker ). A facilitating effect of affirmative sentences was also found by Reuter, Feiman, and Snedeker () in an online comprehension task in which children heard a brief story supported by pictures and indicated their understanding of affirmative and negative sentences by touching an object on a laptop screen. Two-year-old children (mean  months) were able to understand both affirmative and negative sentences when a block of affirmative sentences preceded the negative block, but failed to perform well to either sentence type when negatives were first in the presentation. However, in this experiment, the felicity conditions for negation were satisfied within the discourse in each trial, so the researchers conclude that lack of appropriate pragmatic context in the experimental items cannot have been a factor in the findings. This leads them to conclude that, when presented first, the affirmatives may prime the syntactic structure of the negatives, thus aiding processing of the sentences for this very young group of children. The importance of pragmatic context in comprehending negative sentences is highlighted in an experiment by Nordmeyer and Frank (). In this experiment, - and -year-old children responded to affirmative and negative sentences. There was no preceding discourse. Instead, the pragmatics for the negative sentences were supplied solely by the context in a picture, and children evaluated its felicity on a five-point scale of smiley or notsmiley faces. One group of children evaluated negative sentences such as ‘DW doesn’t have an apple’ in a context in which three characters had an apple but DW had a cat. This context was considered to be informative because there was information about characters who did have apples in the context. Another group of children evaluated the sentence in the uninformative pragmatic context in which the three characters had nothing, and DW had a cat. As one might anticipate, the experimental finding was that children rated the uninformative context significantly lower than the more informative context. In sum, experimental findings are in agreement that pragmatic context is an important ingredient in children’s successful production and comprehension of negative sentences.

.. N 

.................................................................................................................................. In English, negative questions such as What don’t you like? are formed by raising the negative auxiliary verb from I to C, an instance of head movement. Such movement requires the knowledge that a negative auxiliary verb is composed of two heads, Aux+n’t. Therefore, if English-speaking children progress through a stage at which only adverbial forms of negation are available to them, they will be incapable of producing adult-like negative questions, e.g. What don’t you like? Because such wh-questions are derived by raising a negative auxiliary verb from I to C, these wh-questions will become possible only

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

when children have acquired n’t as a head form of negation. Assuming that Englishspeaking children transition to the relevant stage between roughly ages ; to ;, then we anticipate that, by the age of  years old, English-speaking children should produce adult-like wh-questions. Interestingly, adult-like wh-questions with a raised negative auxiliary verb are late appearing in the grammars of -year-old English-speaking children. An elicited production experiment by Guasti, Thornton, and Wexler () found that many -year-old children frequently failed to raise the negative auxiliary verb, preferring to retain negation in the IP. The ten children in the study were between ; and ; years. Only about % of their negative Yes/No and wh-questions were adult-like. The remaining % of children’s questions took one of four nonadult forms. Examples are in () (from Guasti, Thornton, and Wexler : ). The most common nonadult question was the Aux-Doubling structure. This accounted for % of the children’s productions. A Neg-Aux Doubling structure was much less frequent (%). Another strategy was just to leave the negative auxiliary verb unraised (%). Finally, some children favored the Not-Structure illustrated in (d) (%). While this question form is grammatical for adults, children produced such structures in contexts where adults preferred the question form with a raised negative auxiliary. The different nonadult structures produced by children suggests that their grammars are governed by a general strategy to avoid movement of negative auxiliary verbs. () Aux-Doubling a. What kind of bread do you don’t like? Neg-Aux Doubling

(Rosy ;)

b. Why can’t she can’t go underneath? No I to C Raising

(Kathy ;)

c. What she doesn’t want for her witch’s brew? Not-Structure

(Alice ;)

d. Why can you not eat chocolate?

(Darrell ;)

These same children used a range of negative auxiliaries in their negative declaratives and raised the auxiliary from I to C in their positive wh-questions. Children’s failure to raise the negative auxiliary suggests they are being conservative, but why they are so slow to embrace head movement for the negative auxiliary verbs, given they have acquired n’t as a head form of negation, is still to be determined (but see Hiramatsu ). Using an elicited production experiment adopting the same design with Italian-speaking children, Guasti () uncovered a different finding. Italian-speaking children produced four variants of negative wh-questions, all of which are acceptable for adult speakers of Italian. The four varieties of negative wh-questions are illustrated in (): ()

Null Subject a. dove non può mangiare l’arancia? where NEG can eat the orange ‘Where can’t he eat the orange?’

(Clara ;)

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

     Postverbal Subject b. cosa non ta [sa] fare il what NEG can do the ‘What can’t the child do?’ Left-Dislocated Subject c. Luigino, dove non può Luigino, where NEG can ‘Where can’t Luigino go?’ Cleft



bambino? child (Dante ;) andare? go

d. cos’ è che non prende il bambino? what is that NEG takes the child ‘What isn’t the child taking?’

(Davide ;)

(Aurelio ;)

These examples suggest that Italian-speaking children raise the verb along with its cliticized form of negation to C. The fact that these negative wh-question structures are unproblematic for Italian-speaking children suggests that the idiosyncrasies of the auxiliary system in English, and how it is intertwined with negation delay convergence on the adult grammar.

.. D 

.................................................................................................................................. The term ‘double negation’ refers to sentences with two negations, where each negation makes an independent contribution to semantic interpretation. In some structural configurations, the two negations cancel each other out, yielding a positive interpretation. Double negation is sometimes considered to be a peripheral phenomenon that is only triggered by contextual factors (e.g. Larrivée a). One question, then, is whether children can access such interpretations at an early age. To date, this question has only been addressed in Mandarin Chinese and in English. Mandarin Chinese is a typical double negation language (Cheng and Li ). It does not have negative concord interpretations, where two negations yield a single negative meaning. Mandarin has two negative markers, bu and mei(you), and when combined, the two negatives cancel each other out to give a positive meaning. This is illustrated in () (Zhou, Crain, and Thornton : ): ()

a. Ta bu hui bu lai He not will not come ‘He won’t not come.’ b. Mei(you)-ren mei qu canjia shengri juhui. not(have)-person not go join birthday party ‘Nobody didn’t go to the birthday party.’

Previous literature on Mandarin Chinese has shown double negation to be extremely late. Using an act-out task, Jou () found that children younger than  years of age often acted-out a meaning consistent with single negation. However, the task presented the

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

double negation sentences without any contextual support, which would have elevated the difficulty of the task. With these factors in mind, Zhou, Crain, and Thornton () used a Truth Value Judgment task (Crain and Thornton ) to test children’s interpretation of double negation in felicitous contexts. Sentence (b) was one test sentence presented for judgment following a story acted out with toys and props for the child. In the story, three characters get ready to attend a hippo’s party. Two characters get their present wrapped, but Winnie the Pooh has a tooth ache and isn’t able to get ready. The first two characters show up at the party, and pass on the bad news that Winnie the Pooh may not make it. But, at the last minute, Winnie the Pooh shows up with his gift in hand. In such a context, when there is an expectation that Winnie the Pooh won’t be at the party, it is felicitous to produce (b), which is true in the context. Similar false sentences were also included along with controls. Zhou, Crain, and Thornton () tested thirty - and -year-old children. They found that children ; years and over were % accurate on both true and false sentences, while the children younger than ; reached only % accuracy. These children’s responses showed that they tended to interpret the negation as a single negation. This means that children younger than ; either haven’t acquired the law of double negation, or, they have this knowledge but can’t execute it due to limited processing resources, which leads them to ignore one of the negative markers. This is the alternative favored by Zhou, Crain, and Thornton ().

.. D    ?

.................................................................................................................................. English differs from Mandarin Chinese. Although Standard English is considered to be a typical double negation language, many varieties of English permit negative concord interpretations. In fact, some researchers have proposed that English essentially is a negative concord language, but such interpretations are disallowed in Standard English due to sociolinguistic or other factors (e.g. Blanchette , ; Tubau ). This led Thornton et al. () to predict that once English-speaking children have recognized that English has a negative head, they could, in principle, allow negative concord interpretations until such sociolinguistic factors are incorporated into their grammars. In effect, children acquiring Standard English would permit negative concord in the absence of evidence for it in their positive input. To test this hypothesis, Thornton et al. () tested children and adults on sentences like those in (), which are ambiguous between a double negation reading and a negative concord reading. Both readings were made felicitous in the experimental contexts. ()

The girl who skipped didn’t buy nothing a. The girl who skipped bought something. b. The girl who skipped bought nothing.

(Double Negation) (Negative Concord)

To ensure that children are capable of processing sentences with two semantic negations, Thornton et al. included sentences like (), in which one of the two negative markers was

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

    



contained inside a relative clause. This structural configuration does not license either a negative concord or a double negation interpretation. Therefore, even children who assigned a negative concord interpretation to sentences like () would be expected to assign the same interpretation as adults do to sentences like (). ()

The girl who didn’t skip bought nothing

The experiment tested twenty-four - to -year-old children and fifteen adult participants using a Truth Value Judgment task (Crain and Thornton ). On half of the trials, the double negation interpretation was true, and the negative concord interpretation was false; on the other half of the trials, the reverse was true. The main finding was that adult participants produced double negation responses to test sentences like () % of the time, whereas children did so only % of the time. Responses by both groups were similar to control sentences like (), with children responding correctly % of the time, and adults % of the time. A further analysis partitioned the child participants into groups by preferences, where a preference meant that a participant responded in the same way on five of the six trials. On this analysis, fifteen of the twenty child participants exhibited a preference for negative concord interpretations, as did two of the fifteen adults. The authors take their findings to demonstrate that negative concord is, at least initially, part of the core grammar of Standard English.

.. N    

.................................................................................................................................. Negation interacts with other logical expressions in complex ways. Because of these complexities, researchers have been interested in determining when young children, across languages, have mastered the interactions between negation and other logical expressions. Negation is both a ‘licensor’ and an ‘anti-licensor’. As a licensor, negation authorizes the existential expression any in (a). If negation is absent, as in (b), the sentence is unacceptable. ()

a. Ted did not agree with any of the of the reviewers’ comments. b. Ted agreed with *any of the reviewers’ comments.

As an anti-licensor, negation only tolerates a particular interpretation of another existential expression, some. Consider example (a). Here, some is understood to have scope over negation, so the sentence can be paraphrased as follows: There are some reviewers’ comments that Ted doesn’t agree with. Because some takes scope over negation, the sentence remains acceptable if negation is removed, as in (b). Example (b) is the positive counterpart to the negative example (a). ()

a. Ted did not agree with some reviewers’ comments. b. Ted agreed with some reviewers’ comments.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

As examples () and () indicate, the existential expressions some and any are two sides of the same coin. Negation must take scope over any, and negation cannot take scope over some. Adopting current linguistic terminology, any is referred to as a Negative Polarity Item, and some is referred to as a Positive Polarity Item. More generally, we can refer to them both as polarity sensitive expressions. Children’s mastery of the interaction of polarity sensitive expressions and negation has been the topic of much recent research in child language. Polarity sensitive expressions are semantically related to words for disjunction (e.g. English or). Suppose that Ted received three reviews in response to a paper he submitted to a journal. In notifying Ted that his paper had been accepted, the editor maintains the anonymity of the reviewers by referring to them as reviewers A, B, and C. As examples (a) and (b) show, in this situation where the individuals under discussion are known, the meaning of example (a), with the existential expression some, is equivalent to example (b), with the disjunction word or. Both some and or are interpreted as taking scope over negation. ()

a. Ted did not agree with the comments of some of the reviewers. b. There were comments by reviewers A, B, or C that Ted did not agree with.

Examples (a) and (b) show that, in the same situation, a sentence with the existential expression any is logically equivalent to one with the disjunction word or. In these examples, Negation takes scope over any and it takes scope over or. ()

a. Ted did not agree with any of the comments by the reviewers. b. Ted did not agree with the comments by reviewers A, B, or C.

The examples in () and () reveal that the English disjunction word or is like a chameleon, in that it can replace both the Negative Polarity Item any and the Positive Polarity Item some, as long as the relevant individuals have been made explicit in the local context. What this shows is that English or is not polarity sensitive. Having established that negation interacts in complex ways with existential expressions and with disjunction, we will now review the findings from studies of children’s comprehension of negative sentences that contain these logical expressions. The English Negative Polarity Item any emerges early in children’s spontaneous productions. The word any is produced by children as young as ; (Tieu and Lidz : ). This finding was reported by Tieu (, ), who surveyed the transcripts of twenty-six children acquiring American or British English. Each of these children produced fifteen or more utterances with any, and only % of these sentences lacked a licensing expression. Sentential negation was the most common licensor. For example, one child (Abe) produced  instances of any between the ages of ; and ;, and % () of these sentences contained negation (Tieu : ). Children’s knowledge that any requires a licensing expression was investigated experimentally, first by O’Leary and Crain (). These researchers interviewed eleven - to -year-old English-speaking children, using an elicited production task. There were two experimenters. One experimenter acted out short stories in front of the child, and the

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

    



second experimenter played the role of a puppet, who described what it thought had happened in the story. The child’s task was to indicate whether the puppet’s statements were right or wrong. On the critical trials, the puppet’s statements were wrong. On these trials, the first experimenter asked the child to tell the puppet ‘what really happened’ in the story, and this explanation formed the elicited production component of the task. The aim of the study was to elicit sentences from children that contained the Positive Polarity Item something or the Negative Polarity Item anything, to see if children adhered to the licensing conditions on these expressions. As example (a) and (b) illustrate, in one condition the puppet’s statements about the story contained the Negative Polarity Item anything. However, because the puppet’s statements were ‘wrong,’ the child participants were compelled to use positive sentences to explain to the puppet what really happened in the stories. The child’s sentences were not expected to contain a licensor for anything. There were  trials of this kind. Only a single child, on a single trial, corrected the puppet by producing a sentence with anything. This clearly demonstrated children’s knowledge of the licensing condition on Negative Polarity Items. ()

a. Puppet: None of the Ninja Turtles got anything from Santa. Child: No, this one found something from Santa. b. Puppet: Only one of the reindeer found anything to eat. Child: No, every reindeer found something to eat.

In another condition, the puppet produced sentences with something, as illustrated in (). In this condition, the child participants were expected to produced sentences with negative expressions, such as none or not, which license the Negative Polarity Item anything. For the most part, children’s sentences contained anything, as illustrated in (). However, seventeen (%) of children’s sixty-one responses in this condition contained the Positive Polarity Item, something, rather than the Negative Polarity Item anything, as illustrated in (). () Puppet: Every dinosaur found something to write with. Child: No, this one didn’t find anything to write with. ()

a. He didn’t get something to eat. b. None of the dinosaurs found something to read.

(C.E-K. ;) (E.P.;)

In producing sentences with something, children conveyed a meaning that Englishspeakers can only convey using the Negative Polarity Item anything. This finding was taken by Crain () as evidence that children do not initially analyze something as polarity sensitive; in contrast to adults, children initially do not require something to take scope over negation. Rather, something is interpreted in its surface position, even in negative sentences. An experiment by Huang and Crain () investigated Mandarin Chinese-speaking children’s interpretation of the Negative Polarity Item renhe ‘any’, which is governed by the same constraint as English any. As illustrated in () Mandarin renhe is licensed by negation (mei ‘not’).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 

() Xiaogou chi le henduo dongxi, keshi xiaozhu mei chi renhe dongxi. Mr. Dog. ate  a lot food, but Mr. Pig  eat any food ‘Mr. Dog ate a lot of food, but Mr. Pig did not eat any food.’ The experiment tested forty-five -year-old Mandarin-speaking children using the Truth Value Judgment Task. In the story preceding the puppet’s sentence () Mr. Pig and Mr. Dog went to a party. At the end of the party, Mr. Dog was full and happy, whereas Mr. Pig had only eaten a tiny prawn and was still hungry. In this context, statement () is false. The main finding was that children rejected the test sentences % of the time. The final topic is the interaction of negation and disjunction. We saw that disjunction is not polarity sensitive in English. For both children and adults, disjunction phrases are interpreted in situ in negative sentences, resulting in the contrast in meanings exhibited in (). In example (a), the disjunction phrase takes scope over negation, yielding the ‘not both’ interpretation. In example (b), negation takes scope over the disjunction phrase, yielding the ‘neither’ interpretation. ()

a. It was sushi or pasta that Ry did not order. b. Ry did not order sushi or pasta.

(Ry did not order both foods) (Ry ordered neither food)

When example (b) is translated into many adult languages, including Italian, French, Russian, and Japanese, the ‘neither’ interpretation assigned by English speakers does not eventuate. The only possible interpretation for adult speakers of these languages is the ‘not both’ interpretation. This is true even in languages, such as Mandarin Chinese, which exhibit the same basic word order as English. It has been proposed that disjunction words are Positive Polarity Items in these languages (Goro ). As such, disjunction must take scope over negation at the level of semantic interpretation, regardless of its position in the sentence structure. Studies of children acquiring these languages, however, have led to an intriguing finding. Children acquiring all languages appear to initially derive the ‘neither’ interpretation, whether or not disjunction appears in the scope of negation in the surface sentence structure (Crain ; Crain and Thornton ; Geçkin, Thornton, and Crain ; Goro and Akiba ; Pagliarini, Crain, and Guasti ). This finding is attributed to a principle of learnability instructing children to initially derive the interpretation that makes sentences true in the narrowest range of circumstances. Most importantly, the observation that children and adults assign different interpretations to sentences like (b) in one class of languages is evidence that children’s initial interpretation of negative sentences with disjunction is not based on the input from adult speakers. This finding, and indeed many of the findings presented in this chapter, shows that negative sentences are a useful tool in adjudicating between alternative accounts of the nature of child language.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ......................................................................................................................

      ......................................................................................................................

 ˊ    

.. I

.................................................................................................................................. I this chapter, we present an overview of mostly generative-oriented studies that investigate how multilingualism affects the development of negation and its interaction with other syntactic phenomena in adult second language learners and, to the extent that it has been explored, in heritage bilinguals, child simultaneous bilinguals, and child second language learners.¹,² While many early studies of the acquisition of negation by L learners focused on whether these learners replicated the developmental patterns of L children (e.g. Cancino, Rosansky, and Schumann ; Ellis ), later studies argued against a universal sequence of negation for L learners (Clahsen, Meisel, and Pienemann ; Meisel ; Ortega ) and investigated L development of negation from a more self-contained perspective, focusing on how and where cross-linguistic influence occurs in the grammars of multilingual learners whether at the syntactic feature level (Austin, Blume, and Sánchez ; Eubank ; Schwartz and Sprouse ), at the level of semantic interpretation (Grüter, Lieberman, and Gualmini ) or at the interface between syntax and other components (Chung ; Hauser ; Perales ; Unsworth ; Verhagen ). In discussing these studies, our chapter will address the following questions: . How does the development of negation in L2 learners compare to to the acquisition of these structures by child L1 learners and simultaneous bilinguals? . How has cross-linguistic influence been analyzed in the acquisition of negation by multilingual learners? ¹ For an overview of L2 studies of negation that cover sociolinguistic and usage-based perspectives, see Ortega (). ² As we will show, research on child bilingual or multilingual acquisition is scarce when compared to research on adult learners.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ˊ    

. How has the study of the development of negation in early bilingual child acquisition and adult second language acquisition informed some of the most influential theoretical approaches to child bilingualism and second language acquisition, for example the Autonomous Development Hypothesis (Meisel ), the Cross-linguistic Influence Hypothesis (Hulk and Müller ; Müller and Hulk ), the question of access to UG (Eubank ; Schwartz ; Schwartz and Sprouse ; Vainikka and YoungScholten ) by L learners, and the Interface Hypothesis (Sorace ; Sorace and Filiaci ; Sorace and Serratrice )? The article is organized as follows. In section ., we provide a brief introduction to how generative theories of early bilingualism and second language acquisition have evolved as well as some of the major sources of parametric variation in negation across languages that could generate difficulties for bilingual and L acquisition. In section ., we present an overview of a selection of studies on negation in heritage and child L acquisition, in section . an overview of a selection of studies of adult L acquisition followed by final remarks.

.. B  L          

.................................................................................................................................. The comparison of adult L language development with L language development was an important characteristic of second language acquisition studies in its early stages (Cancino, Rosansky, and Schumann ; Clahsen, Meisel, and Pienemann ; Ellis ). Equivalent comparisons are rarely made between the patterns observed in adult/child L acquisition and early bilingual first language acquisition, despite the fact that in both cases the process of acquisition involves another language. However, there are important similarities in how theories of bilingual first language acquisition and L acquisition have evolved and have arrived at similar conclusions on how the bilingual mind acquires two or more languages. In this section, we present an overview of some of the most influential views that have shaped studies on the representation of multiple languages in the mind. Numerous studies demonstrate that bilingual children develop separate and autonomous grammatical representations in their languages from a very early age (De Houwer ; Lindholm and Padilla ; Meisel , ; Paradis and Genesee , among others). At the same time, clear evidence of cross-linguistic influence has been found in the developing grammars of bilingual children (Hulk and Müller ; Müller and Hulk ) especially at the interface of syntax and pragmatics. This type of cross-linguistic influence may be a temporary phase that is later overcome when the bilingual child develops a more sophisticated knowledge of the interaction between grammar and language use in the two languages, but in some contexts it may have long-lasting consequences that result in a differential development of some grammatical properties among bilinguals and monolinguals due to a multiplicity of factors such as frequency of activation (Putnam and Sánchez ), quality and quantity of input (Montrul ; Rothman ) among others.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

     



The issue of cross-linguistic influence at the interface of syntax and other components emerged in adult second language acquisition in the early s and became known as the Interface Hypothesis (Sorace ; Sorace and Filiaci ) shortly after similar research was being conducted in bilingual first language acquisition (Hulk and Müller ; Müller and Hulk ). In the late s and early s, the predominant focus of second language acquisition was on issues such as learnability and accessibility to universal principles in contexts in which parametric values for a first language have already been acquired and are part of the steady state of acquisition (for an overview see White ). The early debates centered on whether the universal principles that underlie the faculty of language are directly available to second language acquirers or are mediated by their instantiation in the first language. While proponents of the Fundamental Difference Hypothesis between L and L acquisition (Bley-Vroman ) and the unavailability of access to Universal Grammar (UG) (Clahsen and Muysken ) interpreted non-target L data as evidence supporting their views, other proposals analyzed non-target forms as a reflection of difficulties in the acquisition of morphology rather than a lack of access to UG (Duffield and White ; White ) or as not revealing of the learner’s knowledge of the appropriate parametric setting universal principles (Epstein, Flynn, and Martohardjono ). As generative studies evolved, especially with the publication of the Minimalist Program (Chomsky ), functional feature values became the locus of cross-linguistic variation. Access to resetting feature values rather than parameter values became the focus of L acquisition studies. Different views were advanced such as the lack of projection of functional projections (Vainikka and Young-Scholten ), the underspecification of functional features at the early stages of acquisition (Eubank ), and the transfer of feature values from the L and the subsequent value resetting (Schwartz and Sprouse ). In this debate, negation featured prominently as a means of testing learners’ acquisition of verb movement triggered by functional features. Studies on the L acquisition of negation shifted their focus from the parallelism or lack thereof between stages of acquisition found in L and L acquisition to how L acquisition of the relative position of negation and verbs is shaped by verbal feature specification (Eubank ; Schwartz and Sprouse ). More recently, research on L negation has moved beyond the feature specification debate. An important part of this change was the introduction of the Interface Hypothesis (Sorace and Filiace ). Nevertheless, research on verb placement and negation has continued (Rankin ) along with more detailed studies of the earlier stages of acquisition of negation and what they reveal about the relationship between the lexicon and syntax (Hauser ; Perales ), studies that focus on the acquisition of relative scope (Grüter, Lieberman, and Gualmini ), and studies that locate difficulties in the L acquisition of negation at the the interface between syntax, semantics, and pragmatics (Verhagen ; Chung ). Within the generative syntax tradition, what parametric variation means in terms of negation has evolved depending on whether negation is viewed mostly as a functional head that projects above or below Tense Phrase (Laka ) or as involving interpretable and uninterpretable features that may be carried by lexical items and/or phonologically null operators (Haegeman ; Déprez ; Déprez and Martinau ; Zejlstra , inter alia). In the s, proposals such as Laka () stated that one of the main differences found across languages regarding negation is whether it projects above or below TP.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ˊ    

In Laka’s view, in languages such as Basque negation is generated above TP and in languages such as English negation is generated below TP, as illustrated in examples (): () a. Basque (negation above TP): Nik ez dut liburu hori irakurri -   book this read ‘I haven’t read this book’ b. English (TP above negation): I have not read this book Under this view, one of the main challenges for bilingual and second language learners of languages with diverging parametric values was being able to project negation appropriately either above or below tense. While this was a matter of controversy in that alternative analyses proposed a universal position for the projection of negation either above Tense Phrase (TP) (Belletti ) or below TP (Pollock ), the relative positioning of negation and tense as functional heads became the focus of several bilingual and L studies included in this article on the relative placement of negation, tense, and verbs (Meisel ; Schwartz ; Eubank ; Schwartz and Sprouse ) and this area of research has continued until recently (Rankin ; Schimke and Dimroth ). The arrival of the Minimalist Program (Chomsky ) opened up the possibility of analyzing negation as a feature that must be satisfied or valued. Currently, most analyses (Zanuttini ; Zeijlstra ; based on original work by Pollock ) assume a functional projection, Negative Phrase, whose head may host a lexical negative element carrying an interpretable negative feature [iNEG]. The NEG head may also enter Agree relations with other negative markers in the sentence allowing for an analysis of sentential negation as well as other negative words such as negative polarity items. In fact, in languages with multiple sentential negative markers, it is assumed that at least one of the overt negative markers carries an uninterpretable negative feature [uNEG] that must be valued and requires the presence of an element with interpretable negative features [iNEG] (Zeijlstra ). Despite these developments in syntactic theory, in multilingual acquisition research, the issue of negation placement relative to verb placement remained at the center of the debate on whether features were transferred from the L at the early stages of acquisition or not. With the exception of researchers that took the strong position of claiming that functional projections are not part of the early representations of second language grammars (Vainikka and Young-Scholten ), the early or late specification of verb features in L acquisition became an important part of the debate in generative second language acquisition theories (Eubank ; Schwartz and Sprouse ) and negation became one of the standard tests for verb placement. When the topic of valuation of features became salient in the analysis of languages that exhibit negative concord (Haegeman ; Déprez ; Zeijlstra ) it opened up the issue of how challenging it is for bilingual children and L learners to keep apart the relevant mapping of interpretable features onto different sets of negative lexical items in two or more languages especially when issues of scope and interpretation rise (Austin, Blume, and Sánchez ; Chung ; Grüter, Lieberman, and Gualmini ).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

     



In the next sections, we present an overview of research conducted on child and adult bilingual acquisition of negation. We have cast as wide a net as is currently available in terms of language pairs investigated.

.. T          L  

..................................................................................................................................

... The acquisition of negation in simultaneous child bilinguals A longitudinal study of the acquisition of negation in simultaneous bilingual children was conducted by Paradis and Genesee (), and examined the development of French and English in bilinguals between ; and ; years of age. Previous research indicated that at early stages, French-speaking monolingual children produce negation pre- and postverbally, as shown in examples () and () from Déprez and Pierce (): les voitures () Pas cherch-er Not look-for- the cars ‘Not to look for the cars’

(Phillipe ;)

tourn-e pas () Ça That turn-  not ‘That does not turn’

(Phillipe ;)

Paradis and Genesee hypothesized that their bilingual participants might produce more postverbal negatives than monolingual children learning English due to cross-linguistic influence from French. They found that this was not the case; bilingual children did not produce more postverbal negation in English than monolingual children. These findings of essential similarity between simultaneous bilinguals and age-matched monolingual children show some similarity with the patterns for heritage bilingual children regarding negation reported by Austin, Blume, and Sánchez discussed in section ...

... The development of negation in child heritage bilinguals Austin, Blume, and Sánchez () examined the acquisition of negation and negative polarity items (NPIs) in a longitudinal study of thirteen bilingual children, ages – (at first interview) who were heritage speakers of Spanish growing up in the U.S. The children spoke Spanish at home with their parents and received instruction in school primarily in English, their second language, with minimal instruction in Spanish. The authors

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ˊ    

compared syntactic development in three areas of grammar in which Spanish and English differ: sentential negation, which occupies a position before an inflected verb in Spanish but after an inflected verb in English as well as negative polarity items (NPIs), expressions with different distributions in Spanish and English, and which must be licensed by negation.³ Austin, Blume, and Sánchez found signs of attrition in the children’s interrogative and negative polarity item (NPI) expressions in Spanish, but not Spanish negation. The contrast between the development of negation and NPIs in Spanish is shown in the examples in (). Whereas the production of target-like productions of negation increased over the course of two years in both languages, the development of NPIs progressed in English but underwent attrition in Spanish. The children’s correct production of negation in Spanish increased from % at the beginning of the study to % two years later. In English, the children produced negation correctly % of the time at the beginning of the study, and % after two years. Examples of the correct production of negation in each language can be seen in (). Negation (correct in Spanish and English): () a. No com-e pastel. Not eat-  cake. ‘He does not eat cake.’ b. He doesn’t like apple pie. The development of NPIs in Spanish showed a strikingly different pattern from the development of negation in both languages. At the beginning of the study, children produced NPIs correctly in % of possible contexts, and two years later in % of possible contexts. In contrast, in English the children produced correct NPI expressions % of the time at the beginning of the study, and % of the time after two years. Examples of NPIs in both languages can be seen in (). In (a), an example of an incorrectly produced NPI in Spanish is shown, where the child uses the phrase ni un (‘not one’) instead of the target ningún ‘none’. Negative polarity items (incorrect in Spanish, correct in English): ni un pastel.4 () a. *No le gust-a Not  like-. not a cake. ‘He doesn’t like any cake.’ Target: No Not

le 

gusta like- 

ningún NPI

pastel cake

b. She said she doesn’t like any bread. Austin, Blume, and Sánchez concluded that negation in Spanish was less vulnerable to attrition than NPIs due to differences between English and Spanish in functional feature strength in NPIs. Such a difference is not present in the case of sentential negation. While ³ Their study also included wh-questions. ⁴ There is one interpretation in which this sentence is correct: he does not like even one pie, but that was not the intended meaning of the utterance. A reviewer suggests a phonological account for this example: the dropping of the intermediate consonants. In any case, the contrast between Spanish and English remains.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

     



these bilingual children do not show similar results in both languages to the children in the Paradis and Genesee () study, we would like to highlight that in both cases sentential negation was acquired in both languages quite successfully.

... The development of negation in child L2 learners Early research on the acquisition of English by adult L learners proposed a universal set of stages for the acquisition of negation. This sequence is shown in (), and begins with the production of preverbal negation. While in L acquisition the early stages involve sentence external negation (Bellugi ; Bloom ; and Wode ) and later stages sentence internal negation, the stages in () proposed for L acquisition are all sentence internal but do involve a developmental pattern as noted for L acquisition of negation: ()

Developmental stages of negation in L learners Stage I: preverbal negation no + V, don’t + V, no + phrase Stage II: preverbal and postverbal negation don’t/doesn’t + V, cop/aux + , no/not + phrase Stage III: target pattern ( following the finite element) present/past distinction, restructuring of unanalyzed forms, use of do-support, position of  after cop/aux (Adapted from Meisel ; based on Stauble , )

For English, the sequence reveals the complexity of how lexical items may be inserted as heads of NegP as inflection and Tense develop. At the stage I, inflection is not fully developed so it tends to be preverbal, possibly as the head of NegP. As inflection evolves and inflected verbs or tense features head T negation, lexical forms of negation may appear in postverbal position with auxiliaries or do support in a position higher than the main verb. Despite the fact that acquisition of negation by L and L learners involves multiple stages, Meisel (: ) noted that in L adults the acquisition of negation is characterized by “considerable variability, not only cross-linguistically but also across learners and even within individuals.” He contrasts these findings with the patterns seen in French-speaking and German-speaking children learning their first languages, for whom the acquisition of negation is relatively rapid and “almost error-free.” Cancino, Rosansky, and Schumann () conducted a study of the acquisition of negation in six Spanish speakers of different ages who had been learning English as an L for one to four months:  young children (age  and age ),  older children (age  and age ), and two adults. The authors found that all their participants passed through similar stages in acquiring negation with third person subjects, seen in (), which resemble the stages from Meisel () shown in (). ()

Stage : negation + verb Stage : unanalyzed ‘don’t’ + verb or aux Stage : aux + negation Stage : analyzed ‘don’t’

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ˊ    

These findings reinforced the notion that there are universal stages of acquisition of negation although the question of what triggers them remained open for debate as it involves a more detailed analysis of the relation between syntax and morphology in L development. Negation interacts not only with Tense; in some languages like Dutch, it interacts with phenomena such as object scrambling. Unsworth (, ) examined the acquisition of direct object scrambling in L Dutch by child and adult learners who were native speakers of English learning Dutch as a second language. In Dutch, direct objects can be moved from their postverbal position to the left of negation for pragmatic effects such as focus and topicalization (Reinhart ; Neeleman and van de Koot ). Learners must acquire scrambling and maintain the scope of negation over the scrambled object. () a. Jan heeft niet [de boom] geplant (non-scrambled) John has not the tree planted b. Jan heeft [de boom]i niet ti geplant (scrambled)5 John has the tree not planted ‘John didn’t plant the tree yesterday.’

(Unsworth, )

The participants in Unsworth’s study were twenty-five child L learners who had begun acquiring Dutch between the ages of ; and ; years, and were between  and  years old (mean ; years) at the time of testing. She also tested twenty-three adult L learners who were first exposed to Dutch between  and  years old, and were  to  years old at the time of testing (mean ; years). Unsworth divided the participants by proficiency so that there were three groups of child L learners and three groups of L adults. She found that there was no statistical difference between the child and adult L learners in their use of scrambling, and that both groups passed through the same stages at different proficiency levels: the low proficiency children and adults failed to scramble objects, whereas the mid and high proficiency groups did so consistently. Failure to acquire scrambling reflects an inability to move elements outside the VP and above NegP but below TP. In discussing the results from Cancino, Rosansky, and Schumann and earlier data from Unsworth (), Schwartz () interpreted these results as support for the hypothesis that adult and child L learners have access to Universal Grammar, as well as for the Domain by Age Model (DAM), which proposed that the acquisition of syntax by child L learners resembles that of adult L learners, whereas children’s L acquisition of inflectional morphology is similar to its development in child L learners (Schwartz ).⁶ The extent to which negative items such as don’t in English that are syncretic with tense and agreement are analyzed as chunks or as involving morphological marking of functional features at the early stages of non-adult second language acquisition has also been studied from the perspective of second language acquisition in bilingual first language acquirers.

⁵ Unsworth’s () task is based on Schaeffer (). It elicits negative sentences with a previously introduced topic. ⁶ The acquisition of morphology has been shown to be particularly challenging for adult L2 learners as it involves cross-linguistic differences in mappings of syntactic features onto morphological forms (Prevost and White ; Lardiere ; Slabakova ; among others).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

     



In a study on the acquisition of negation by Basque-Spanish bilingual children ages – and teenagers aged –, Perales () compares negative sentences produced by BasqueSpanish learners of English as a second language with those of L acquirers from a previous study by Schütze (). She adopts as a premise Schütze’s proposal according to which when Tense and/or agreement are underspecified two forms emerge not and don’t. When Tense is underspecified () is the case and when agreement is underspecified () obtains. Both forms correspond to third person singular subjects. () ()

not+V (e.g. he not go) don’t + V (e.g. he don’t go)

Perales’s results show that despite having had more years of exposure to English the teenagers and the younger children have similar percentages of non-agreeing don’t with third person subjects, which could lead to the interpretation that negation is acquired while agreement remains underspecified. Perales’s findings, however, also indicate that there is a significant difference between the younger children’s errors and those in the older group in the production of negation. The younger L Spanish learners use don’t in contexts in which forms such as isn’t are required as shown in () more frequently than those in the older group and in stark contrast with L acquisition data. Perales notes that these types of errors have not been attested in L acquisition data and she takes these facts to indicate that rather than being an underspecified negative item, don’t is actually used as a lexical chunk that remains unanalyzed. ()

and the frog doesn’t (isn’t) in the house

(Perales ; )

We would like to argue that doesn’t as a lexical item in sentence () might be unanalyzed with respect to the auxiliary verb but the learner is showing the correct morphological mapping of negation, tense, and agreement features. Overall, while there is some evidence of similar patterns of acquisition in bilingual children, child L, and adult L learners, some important differences emerge with respect to NPI acquisition among heritage bilinguals as well as in syncretic forms involving negation among older child L learners. Some of these differences will be noted again in our overview of adult L acquisition.

.. N   L 

.................................................................................................................................. In this section we will provide an overview of some studies on adult second language acquisition of negation that includes studies that aim to provide an account of the early stages of negation, studies that focused on the debate on the resetting of functional values based on negation and verb placement data, and more recent studies that focus on the interpretation of negation at the interface between syntax and the lexicon and syntax and semantics.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ˊ    

... Formulaic negation Hauser () analyzes, from a conversational analysis perspective, the oral production of Nori, an adult learner of English as a second language whose first language is Japanese and who lives in Honolulu at the time of the study. Hauser traces the development of negation in the speech of this L learner from early stages of constituent negation to the productive use of don’t in a longitudinal study over a thirty-week period. Overall, Nori’s production of what Hauser labels no(-t)-X forms (, ) and X-no(t) forms () is stable. () No understand ()

If . . . no busy.

()

Today no

(Hauser : )

While X-no(t) negation has a wide range of uses, X-no(t) negation is restricted to corrections in Nori’s speech. The X-no(t) expression in () is used in a context in which Nori corrects his own previous use of the word today that has lead to a misunderstanding in the conversation. While these two types of negation appear to be stable throughout the recordings, Hauser points out that Nori also makes productive use of the negative formula I don’t know in his interactions with his interlocutor in the recordings and gradually progresses from using don’t only with the verb know to using it with the verb like. Nori also moves from using the expression only with the first person pronoun I to using it with the second person pronoun you sometimes as a repetition and sometimes as a correction. Hauser highlights the relevance of idiomatic expressions containing negation as providing a resource for the development of negation for some L learners. He concludes that the results of his longitudinal case study can be taken as providing support for a usage-based learning view of adult second language acquisition. We would like to take a different position and analyze the data produced by Nori as showing first a stable projection of a pre-predicate no(t)-X negation and the availability of some form of topic position for the predicate in the cases of X-no(t) for correction purposes. The pre-predicate negation corresponds to a stage in which the negative element is the head of NegP. Before the development of inflection, it merges with a Predicate Phrase. Since Japanese is a language with generalized fronted topics (Vermeulen ), it is not surprising that a fronted topic could be merged with NegP and an elided Predicate Phrase resulting in (). In our view, these data are compatible with a representational analysis such that there is an early stage of merge that may or may not include the specification agreement or tense, but consistently shows differences in the relative position of the verb and negation. Clearly, topic-neg order cannot be present in the input Nori receives from his native English speaker interlocutor. In that respect, evidence in favor of a lexical reanalysis of the negation within the idiom I don’t know does not necessarily imply that only a usage-based approach can account for the facts, as an early syntactic representation of negation can coexist with the lexical reanalysis of the idiom.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

     



... Negation as a test In the ’s the debate in generative approaches to second language acquisition focused on access to Universal Grammar. As mentioned above, two competing analyses on feature specification were put forth that focused on feature values. Eubank’s (/, ) Valueless Features Hypothesis characterized the early stages of second language acquisition as a period in which functional features remained unspecified. On the other hand, Schwartz and Sprouse’s (, ) Full Transfer/Full Access Hypothesis proposed an initial stage of acquisition characterized by the transfer of the value specifications from the L to the L with subsequent resetting of value specifications which they took as evidence of access to UG. An important aspect of this debate in the ’s centered around the issues of the relative placement of verbs and negation that continues to be at the core of some recent work on the L acquisition of negation. One of the first steps in assessing that the relative placement of verbs and negation was revealing of access to UG was the work by Tomaselli and Schwartz () on the acquisition of verb placement in L Romance-L German. They reanalyze the three stages of negative placement in the L German, a V- language, of L speakers of Italian and Spanish, both languages with canonical SVO word order in main clauses presented by Clahsen and Muysken ().⁷ The three stages of negation are preverbal negation, postverbal negation, and Trennung (separation of negation and the verb) which in their view roughly coincide with the six stages of verb placement in the German interlanguage. The sequence is shown in Table . These authors analyze the first stage of negation (many details are omitted due to space limitations) as involving Neg as an adjunct to VP and a lack of V-to-I movement only for lexical verbs but not for auxiliaries, as seen in () and (). In the second stage of negation there is V-to-I and they assume headedness of the VP has changed to head-final (). In the third stage of negation, based on a series of assumptions about directionality of adjunction, they propose direct object scrambling as a form of adjunction to the left of VP.

Table .. Stages of negation and verb placement in L Romance-L German i)

preverbal negation

ii) postverbal negation iii) NEG Trennung

i) ii) iii) iv) v) vi)

S (AUX/MOD) V (O) (AdvP/PP) S (AUX/MOD) V (O) S V[ +fin] (O) V[- fin] XP V[ +fin] S . . . V[ +fin] (AdvP) O V[ +fin] clause-finally in embedded clauses

Based on Tomaselli and Schwartz (1990)

⁷ VP-internal subjects in Spanish raise to the Specifier of TP (Zagona ).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 ()

 ˊ     Nich sprechen Italien Italian not speak (Tomaselli and Schwartz : )

() Die familie is nich gekomm von Spanie from Spain The family is not come ‘The family didn’t come from Spain’ (Tomaselli and Schwartz : , from Clahsen : ) ()

Aber er hat nich die papier But he has not the paper ‘But he hasn’t got the papers’ (Tomaselli and Schwartz : , from Clahsen : )

In this early analysis negation is analyzed as an adjunct without the projection of a NegP, unlike in more recent analyses. In later work, Schwartz and Sprouse () provide more evidence for the view that in the process of L acquisition UG is accessible and that feature values can be reset so that they trigger V-to-I-to-C movement in the L German of an L speaker of Turkish, a language with canonical SOV word order. They also argue against Vainikka and Young-Scholten’s () Minimal Trees Hypothesis and against Eubank’s () Valueless Hypothesis. In the first case, they argue against the idea that L learners lack functional projections by showing that L French–L English learners show some evidence of V-to-I in their L. They also argue against the notion that this is due to an overregularization based on the pre-adverbial and pre-negative position of auxiliaries in English found in the input. Schwartz and Sprouse point out that the supposed overregularization does not extend to lexical verbs in L English data. Their argument against Eubank’s hypothesis, according to which lexical and functional categories as well as their linear orientation transfer from the L to the L but their feature strength does not, is also based on their reanalysis of variable verb placement with respect to adverbs and negation in the L English of French speakers. Based on the absence in White’s data (/, , a, b) of cases in which negation follows a thematic verb (unexpected under the Full Transfer/Full Access Hypothesis), Eubank () proposes that the strong feature that triggers verb raising in French has not been transferred to L English as no evidence of postverbal negation is found in the interlanguage; instead preverbal negation is found: ()

I not love you

Schwartz and Sprouse () argue against this proposal on the basis of their analysis of verb placement in questions, their placement relative to sentence-internal adverbs, and negation. Of concern to us is their argument regarding negation. They question Eubank’s analysis of English not in the interlanguage as the counterpart of French pas. In their analysis not in the L early stages of acquisition corresponds to the negative clitic ne in French but unlike in Tomaselli and Schwartz’s () study, negation is no longer an adjunct but a functional head. Not is the head of Neg and its morpho-syntactic properties

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

     



are transferred from the L so that like ne it cliticizes onto the verb, resulting in preverbal negation in the interlanguage. As we have seen, the debate on the L acquisition of negation was closely linked to the acquisition of verb placement. More recently, L research has moved beyond the debate on access to UG expressed by access to a non-native value specification of features and into the realm of the difficulties that the acquisition of syntactic phenomena at the interface of syntax and other language components pose to L learners (Sorace and Filiaci ; Sorace ). However, given the previous debates on word order, negation, and feature specification involved verb placement, even recent studies have continued to look at negation mostly as a test for verb placement (Schimke and Dimroth ) without focusing on the properties of negation itself. At the same time, other studies have emerged that explore areas in which negation is part of the core syntax or it lies at the interface between syntax and other language components. Rankin’s () study corresponds to the first trend. It investigates the extent to which verb-second syntax (V) from L German and Dutch is transferred into L English from the perspective of the Interface Hypothesis. He analyzes corpus data from L German, L Dutch, and L French learners of English, the latter as a control for V effects. His findings indicate that, while there is transfer of V patterns in auxiliary inversion and copula inversion among L German and L Dutch speakers, V is not produced with sentential negation. In fact, there are fewer than % of cases of non-target thematic verb placement with sentential negation. Rankin analyzes these results as evidence in favor of L transfer only in cases where the interface between syntax and pragmatics is involved supporting the Interface Hypothesis (Sorace and Filiaci ). The V strategies used by English L learners in affirmative declarative sentences seem to be driven by pragmatic considerations such as what can be topicalized in the L but not in the L as in () and (), both of which are not acceptable in conversational registers in English: ()

Important for today is the positive acknowledgment of the each opposite sex’s qualities. Essential is just who decides what we can watch and why. (ICLE-GE⁸)

()

Ironic is Jim’s remark about this. Striking is the absence of self-esteem in the black community in the story. (ICLE-DU)

Crucially for this analysis such cases were not found in the L French–L English data, a fact consistent with differences in topicalization strategies between the V languages and French. Given these results, in Rankin’s analysis, the L acquisition of sentential negation falls within the core syntax and not at the interface of syntax and other components. A closer approach to how the syntax and the semantic value of negation interact at the initial stages of L acquisition can be found in Verhagen (). This study investigates the acquisition of postverbal negation and postverbal adverbs in L Moroccan Arabic–L

⁸ The acronym corresponds to International Corpus of Learner English. GE is German L1 and DU is Dutch L1.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ˊ    

Dutch and L Turkish–L Dutch learners. In Dutch sentences with inflected main verbs, negation is postverbal: ()

Hij werkt niet in Amsterdam. He works not in Amsterdam ‘He does not work in Amsterdam.’

(Verhagen , )

Moroccan Arabic and Turkish differ from Dutch with respect to the relative placement of negation. While in Moroccan Arabic negation is expressed via a discontinuous morpheme involving a prefix and a suffix () and both SVO and VSO orders are possible, in Turkish negation is marked only with a suffix and the verb appears in final position as in (): ()

Ma-ka-jakol-. Neg-dur-eat-.- ‘He does not eat.’

oku-ma-dt. () Hasan kitab-t Hasan book- read-- ‘Hasan did not read the book.’

(Verhagen : , from Harrell : )

(Verhagen :  from Kornfilt : )

Verhagen () analyzed the elicited oral production data from a film-retelling task and a picture story task looking at the distribution of negation as well as different types of adverbs, and inflected and non-inflected verbs produced as main verbs. Results show higher percentages of preverbal negation than preverbal adverbs among the L Moroccan Arabic speakers but a higher percentage of preverbal adverbs and negation among L Turkish speakers. The first group showed significantly lower levels of preverbal negation (%) than postverbal negation (%). The difference was not as stark in the Turkish group (% preverbal vs. % postverbal), possibly due to the fact that, unlike Moroccan Arabic that has either sentence-initial or sentence-final adverbs, Turkish has a non-sentence-initial position for adverbs between the subject and the object that is preverbal. This preverbal position for adverbs in Turkish may lead L learners to overtly express a negative lexical item in a preverbal position in Dutch. In order to account for what triggers the acquisition of negation, Verhagen argues that finiteness plays a role in the L acquisition of negation. At early stages of L acquisition of Dutch, non-finite forms appear preceded by negation in topic-predicate structures as in: () Dames niet goeje chauffeur. Ladies not good driver ‘Ladies are not good drivers.’ (Verhagen :  from Jordens and Dimroth : ) Following Jordens and Dimroth (), Verhagen () claims that once finiteness appears on the verb, negation ceases to be what they term “a linking element” between the topic and the predicate. At that point, finiteness takes over the role of expressing the assertion and negation appears in postverbal position. This study exemplifies an attempt to

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

     



locate the L acquisition of negation in its earlier stages at the interface between syntax and informational structure. As part of the shift of focus from functional features to the interface of syntax and other language components, some researchers have begun to explore the difficulties posed to the acquisition of L syntactic properties by discourse-pragmatic properties associated to them. In the case of negation, its interaction with the interpretation of quantifiers and their relative scope has begun to be explored as a way of testing how negative features interact with semantics and pragmatics. This is an attempt to go beyond the view of second language acquisition as involving only feature values. Chung () investigated the interpretation of the quantified expression every NP in object position in negative sentences among L-Korean L-English adult learners. It is a well-known fact that in Korean, child and adult speakers have a strong preference for the interpretation in which the quantified expression has scope over negation. This interpretation has been attributed to processing of the canonical SOV word order in the language (O’Grady ) with short form negation () and long form negation (): motun salam-ul an-manna-ss-ta () Chelswu-ka Chelswu- every person- -meet-- ‘Chelswu did not meet every person.’ motun salam-ul () Chelswu-ka person- Chelswu- every ‘Chelswu did not meet every person.’

manna-ci meet-

(Chung : ) anh-ass-ta  do-- (Chung : )

In a previous study with non-contextualized online and offline tasks (Lee ), advanced L learners exhibited a strong Korean-patterned preference in the offline task and no preference in the online task. Chung hypothesized that there are three possible causes for difficulties in the acquisition of scope in these sentences: L transfer of syntactic computations, inability to integrate syntactic and semantic computations, and lack of pragmatic development in the L. To test this hypothesis, Chung used a contextualized offline acceptability judgment task that presented learners with two possibilities: one in which the quantified expression had scope over negation and another one in which negation had scope over the quantified expression with sentences such as: () Mary didn’t eat every meal

(Chung : )

In one type of context the interpretation would be one in which Mary skipped some meals (Neg > every NP, surface scope). In the other one, for every possible meal it was the case that Mary did not eat it, namely she went hungry all day (every NP > Neg, inverse scope). If L transfer is the main factor affecting L development, then a strong preference for inverse scope was expected across proficiency levels. If L learners can derive both meanings from syntactic-semantic computations, but there are difficulties with interpreting the context (pragmatics), then there would be no preference for either scope. Chung’s results show that advanced learners and the native controls had high acceptance of surface scope while intermediate learners showed more acceptance of inverse scope. Unlike native speakers,

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ˊ    

advanced learners showed more variability across individuals and some ambivalence in their judgments evidenced by the fact that they assigned similar scores to both interpretations in the – scale. Chung interprets these findings as indicative of difficulties among adult L learners integrating pragmatic information despite possessing mature cognitive mechanisms and pragmatic abilities in their L. Despite the growing trend to account for difficulties in L acquisition as stemming from the complex integration of syntax and other components, we end this overview with a study that focused on the scope of negation without testing the Interface Hypothesis. Grüter, Lieberman, and Gualmini () conducted two experiments, one with L English–L Japanese learners and another one with L Japanese–L English learners. They tested them on their interpretation of sentences such as: ()

John does not speak English or German.

and doitsugo-ga hana-se-nai () John-wa eigo-ka John- English- German- speak-can- ‘John does not speak English or does not speak German.’ While the English sentence is interpreted with negation as having scope over the disjunction, the sentence in Japanese is interpreted with disjunction having scope over negation. In other words, () is only true if John does not speak both languages while () can be true if John speaks one of the languages and if he does not speak both. This type of interpretation is late acquired in L Japanese. Based on these facts, Grüter, Lieberman, and Gualmini hypothesize that L English–L Japanese learners will be able to acquire wide scope of disjunction over negation in their L in the same way L Japanese children do. However, the L English of L Japanese learners would show clear evidence of full transfer as their interpretation is a superset of the target interpretation. Their results support their hypothesis for this group. We would like to point out that Grüter, Lieberman, and Gualmini consider this to be a strict case of semantic transfer not involving difficulties in the integration of semantics and pragmatics. However, in our view it constitutes a case of difficulties at the interface of syntax and semantics.

.. C 

.................................................................................................................................. The study of the acquisition of negation among child bilinguals and adult L learners, although not as extensive as the study of other areas of syntax, has nonetheless yielded significant results for theories of bilingual and second language acquisition. While there is little research on the acquisition of negation by multilinguals who are not adult L learners, that is simultaneous bilinguals, heritage bilinguals, and child second language learners, the few existing studies have found similar patterns in the development of negation in simultaneous bilingual children acquiring French and English and age-matched monolinguals (Paradis and Genesee ) providing support to the Autonomous Development

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

     



Hypothesis. Furthermore, despite the lack of evidence of attrition of sentential negation in the Spanish of child heritage speakers learning English as a second language, NPIs in Spanish did show attrition (Austin, Blume, and Sánchez ) supporting cross-linguistic effects. In research on child L acquisition of object scrambling, Unsworth (, ) found no significant differences between the developmental stages of child and adult learners of Dutch as a second language who had been matched for proficiency. Perales () compared the acquisition of negation in English as an L by two groups of Spanish/Basque-speaking children, a younger group age – and teenagers between  and  years old. While both groups of children in Perales’s study produced non-agreeing don’t at similar rates, the younger group also extended its use to contexts requiring isn’t showing evidence of appropriate expression of tense and negation despite the inadequacy of the auxiliary. Early research on the development of negation in L learners focused on the degree to which UG constrains L acquisition as well as the ability of adults to acquire a second language in a native-like way. These studies led to comparisons between stages in the development of negation in L adults and L children (Cancino, Rosansky, and Schumann ; Ellis ) that show similarities in stages between L and L acquisition of sentenceinternal negation. The focus in subsequent years shifted away from the question of whether there is a universal sequence of stages in acquiring negation for L learners towards explanatory factors in the non-target production of negation in L adults, such as crosslinguistic influence in syntactic feature specification (Eubank ; Schwartz and Sprouse ). This resulted in research showing evidence for initial transfer of morpho-syntactic properties of negation from the L to the L (Schwartz and Sprouse ) although in some studies, the syntax of negation was used as a diagnostic for verb movement in order to investigate properties of the L grammar, rather than a focus of research in itself. More recent research has investigated the difficulties that L learners face in acquiring negation, and has proposed that these problems stem from the involvement of interfaces between the syntax, semantics, and pragmatics (Chung ; Hauser ; Perales ; Rankin ; Verhagen ). These studies found support for the Interface Hypothesis, which claims that interfaces between syntax and external cognitive domains such as pragmatics are inherently difficult for L learners (Sorace and Filiaci ; Sorace ).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ........................................................................................................................

EXPERIMENTAL INVESTIGATIONS OF NEGATION ........................................................................................................................

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ......................................................................................................................

  Issues in the processing of negation ......................................................................................................................

    

.. I

.................................................................................................................................. N has played a major role in early psycholinguistic research. Studies here focused on the question of whether, and if so why, negation is more difficult to process than affirmation. This research presupposed that meaning representations created during language comprehension are propositional in nature and that negation is represented explicitly in this propositional format. Research up to today continued addressing this issue of processing difficulty of negation, but various additional questions arose. Also, there was a general shift from propositional accounts of meaning representation to situation-model or mental-model theory (in the s; e.g. van Dijk and Kintsch ; Johnson-Laird ) and then to the experiential simulations view of comprehension that was proposed in the context of the embodied cognition framework (e.g. Barsalou ; Zwaan and Madden ). Thus, research on negation processing over the years took place against the background of changing representational assumptions. In fact, negation proved to be an ideal test case for many of the basic representational assumptions in language comprehension research. In the following, we will first report on early investigations into the processing of negation in section . and will then take this as a starting point for discussing more recent negation-related research. Sections . to . will be organized according to the main issues that are currently debated in this field.

.. E      : I        , ?

.................................................................................................................................. Negative sentences came into the focus of psycholinguistic investigations in the s. Researchers were particularly interested in the difficulties that negation poses during

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



    

processing. Some studies dealt in particular with the impact of negation on the maintenance of sentences in memory. Most studies in those years however focused on the processes involved in the comprehension and production of negative sentences.

... Memory for negative sentences One motivation for examining the memory for negative sentences was the goal of investigating the psychological reality of Chomsky’s () transformational grammar. The hypothesis was that people would capture in memory the deep structure of a given sentence together with the transformations required to achieve the surface structure of that particular sentence. If so, negative sentences should be associated with a higher memory load than the corresponding affirmative sentences as they always involve one extra transformation (the one that transforms an affirmative sentence into its negative counterpart). This led to the prediction that negative sentences should be more difficult to memorize, and that negative sentences should sometimes be falsely memorized as affirmative sentences. The first prediction was indeed borne out, both in studies with free recall (Boysson-Bardies ; Engelkamp and Hörmann ) and in studies with cued recall of previously studied sentences (Engelkamp, Merdian, and Hörmann ; Hörmann ; for recognition studies, see Fillenbaum ; Kurcz ). The second prediction was not supported. Negative sentences were not accidentally recalled as affirmative sentences, but errors with negative sentences rather reflected the fact that in some languages there are several possible positions for the negation operator in the sentence. Thus, negative sentences were often recalled as negative sentences with the negation operator appearing at a different position than in the original sentence. Consequently, the results of these studies were generally taken as evidence against the psychological reality of transformational grammar. Interestingly, some of these early studies also looked into pragmatic factors influencing the memorability of negative sentences. The basic assumption was that negative sentences should be easier to memorize when they are pragmatically felicitous, and that pragmatic felicity is tied to the ease with which an alternative state of affairs comes to mind (Hörmann ). In line with this prediction, negative sentences such as The river did not erode the bank with an inanimate subject and a close relationship between subject and object proved to be harder to memorize than sentences such as The dog did not steal the bone with an animate subject and a looser relationship between subject and object (Engelkamp and Hörmann ; Engelkamp et al. ).

... Comprehension and production of negative sentences Several studies with quite different methodologies investigated the impact of negation on the comprehension and production of sentences. Most of these studies employed verification paradigms in which participants evaluated the truth of a given sentence with respect to a presented picture or their background knowledge. Other studies used sentencecompletion paradigms or investigated the impact of negation even more indirectly by examining the ease with which participants could draw inferences or select target entities after reading affirmative vs. negative sentences (see Table . for an overview). In all of

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 



Table .. Overview of early studies investigating the processing of negation. Sentence-picture verification Gough ,  Clark and Chase ,  Just and Carpenter  Carpenter and Just  Trabasso et al.  Chase and Clark 

The boy [hit] / [did not hit] the girl. The star [is] / [is not] [above] / [below] the plus. The dots [are] / [are not] red; [Many] / [Few] of the dots are red. It’s [true] / [not true] that the dots [are] / [are not] red. [LUZ] / [KOV] green. The circle is [present] / [absent].

Verification against background knowledge Arroyo  Wason ; Wason and Jones  Eiferman  Wales and Grieve 

Miami [is] / [is not] an [American] / [French] city. [] / [] [is] / [is not] an even number; [] / [] [MED] / [DAX] an even number. Seven [is] / [is not] even. Given  and  the next number [is] / [is not] [] / [].

Sentence completion (missing word in grey) Wason  Wason  Wason  De Villiers and Flusberg  Donaldson 

There is [both] / [not both] Yellow in  and Red in  [Six] / [Seven] [is] / [is not] an even number. Circle No  [is] / [is not] blue. There [is] / [is not] a flower. The circle [is] / [is not] red.

Drawing inferences from negative sentences Carpenter  Just and Clark 

[Mary stayed since Judy lived] / [Mary would have stayed if Judy would have lived] ! Mary [stayed] / [left] and Judy [lived] / [died]. John [remembered] / [forgot] to let the dog out ! the dog is supposed to be out.

Following instructions for entity selection Jones a, b; Jones  Donaldson 

[Mark the numbers ,,,] / [Mark all numbers, except/but not , , , ] It’s [black and a circle] / [not black and a circle] / [not black and not a circle]

these studies, negative sentences were associated with longer processing times and/or more errors than the respective affirmative sentences. Considering the breadth of the different types of negative sentences investigated and of the implemented experimental paradigms, it seems that processing difficulties due to negation (in the following termed “the negation effect”) is a very stable phenomenon. Negation is difficult not only with explicit negation markers in the sentences but also with more implicit negation (e.g. forgot, absent, few) or when senseless syllables with the same function are used (e.g. ‘DAX’, ‘KOV’). In both of these latter cases, however, the negation effect tends to be smaller than with explicit

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



    

negation (Clark ; Jones ; but see Sherman ; Wason and Jones ). Furthermore, the negation effect shows when negative sentences are less specific than their affirmative counterparts but also when participants can transform the negative sentences into affirmative sentences with the same meaning (e.g. transforming ‘not open’ into ‘closed’ or transforming ‘not left’ into ‘right’ when the context only involves these two alternatives, cf. Dudschig and Kaup ). In addition, in sentence-picture verification studies, the negation effect is observed both when the picture is presented with or after the sentence and also when the sentence is presented prior to the picture (Trabasso et al. ). In contrast to the rather clear results concerning the impact of the negation operator, the impact of the truth value of the sentences (i.e. whether they are true or false) was ambiguous. In some studies, a main effect of truth value was observed, with true sentences being easier to process than false sentences independent of polarity (affirmative vs. negative; Arroyo ; Eiferman ; Gough ; Just and Carpenter ; Trabasso et al. ; Wason , ). In other studies, a truth-value-by-polarity interaction was observed, with true affirmative sentences being easier to process than false affirmative sentences but false negative sentences being easier to process than true negative sentences (Arroyo ; Carpenter and Just ; Chase and Clark ; Clark and Chase , ; Eiferman ; Gough , ; Just and Carpenter ; Trabasso et al. ; Wales and Grieve ; Wason and Jones ). A lot of research effort was put into developing models that would explain the observed patterns. To account for the two patterns of verification latencies, it was suggested that comprehenders encode the pictures/their background knowledge, just as the sentences, in a propositional format (e.g. even[six]; not[even[seven]]). Both representations are then compared constituent by constituent, whereby an internal response parameter is changed from true to false and vice versa each time an incongruity is being detected. Each incongruity prolongs the time needed to verify a sentence (for a detailed description of the model, see Carpenter and Just ; Clark and Chase ). Two strategies can be distinguished that produce the two observed response time patterns. The first strategy is to use the original sentence representation for the comparison process, and this leads to a truth-value-by-polarity interaction. The second strategy leads to a main effect of truth value. Here comprehenders convert the negative sentences into affirmative ones with the same truth conditions before the comparison process is started (e.g. Six is not an even number is converted into Six is an odd number). Despite the many studies in which negation and truth value were being manipulated, the debate about the definite criteria for when participants employ one or the other strategy is still not settled. We will come back to this issue in section .. The question that arises from the studies reported in this section is why negation is so difficult that it on the one hand consistently prolongs processing times and on the other hand leads participants to adopt processing strategies that allow getting rid of the negation early on during processing. Different explanations have been discussed. First, advocates of transformation grammar assumed that difficulties arise because of the additional transformation that is required for arriving at the deep structure. However, as mentioned above, this type of explanation was ruled out early on by the results of several studies on memory for negative sentences (e.g. Engelkamp et al. ), and we will therefore not discuss it any further. Second, negation may lead to processing difficulties because of its unpleasant connotation reflecting its strong association to prohibition, especially in early childhood. The above-mentioned findings that implicit negation leads to smaller negation effects than

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 



explicit negation (Clark ; Jones ), as well as the results from studies using senseless syllables (e.g. KOV instead of an explicit negation marker; Wason and Jones ), are in line with this hypothesis because in both cases the negation marker that presumably is associated with prohibition is not explicitly mentioned in the sentences. Additional evidence comes from a study by Eiferman () showing that in Hebrew the negation particle ‘lo’ which is used for prohibition led to higher latencies and more errors compared to the negative particle ‘eyńo’ which is only used in denials. A third reason why negation may be particularly difficult was first suggested by Wason (). He suspected that the negation effect might reflect an experimental artefact because negative sentences are often presented without a context in experimental studies. His argument is based on the assumption that in real life negative sentences are usually used to indicate deviations from expectancies or to correct a false proposition (cf. Givon , ). When presented outside of an adequate context, participants may therefore accommodate these expectancies or false propositions as part of comprehending the negative sentence. In other words, when processing a sentence such as Six is not an odd number the comprehender would also consider Six is an odd number because this is the false proposition or expectancy that the negative sentence is correcting. Thus, according to this reasoning, negative sentences are not inherently more difficult to process than affirmative sentences, but only when presented without a legitimizing context. A number of experimental results provided by early negation research was indeed in line with the assumption that negation is easier to process when it is used in an adequate context. For instance, in a study by Wason (), participants were presented with pictures of eight circles, seven in one color and one in another color. They were then asked to complete affirmative or negative sentence fragments. Negative fragments (e.g. Circle No  is not . . . ) took longer to complete than affirmative fragments, but this difference was significantly smaller when the negative sentences referred to the circle with the exceptional color than when they referred to one of the seven other circles. Thus, participants profited from an adequate context when processing negative sentences. Similarly, in a study by de Villiers and Flusberg (), the ease with which a sentence such as This is not a . . . could be completed by preschool-aged participants depended on the similarity between the target and the competitor objects, and in a study by Watson () negation was preferably used when referring to an entity that differed from other entities in terms of a missing property (e.g. a horse without a saddle in the middle of horses with saddles). In addition, in a study by Johnson-Laird (cited in Wason and Johnson-Laird ) participants were asked to paint strips of papers with blue and red crayons to match sentences such as The left hand end of the strip [is red]/[is not blue]. More blue was used for the negative compared to the affirmative sentences, which fits well with the assumption that the negative sentence implies a context in which the strip could be falsely viewed as being all blue. Finally, in a study by Cornish () participants were asked to complete a sentence such as This circle is not all . . . by pressing a blue or a red key as fast as possible. Participants more often used the color that was predominant in the circle even though any answer would have rendered the sentence true. Of course, in both of the latter studies it cannot be ruled out that participants simply chose the respective color because it was salient in the context. We will come back to the issue of pragmatic aspects of negation processing in section . as it is still one of the main topics investigated in present-day negation research.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



    

To summarize, early studies on negation processing have focused on the question whether, and if so why, negation is more difficult to process than affirmation. The results of this research impressively demonstrate that negation indeed leads to processing difficulties across a broad range of sentence types, experimental tasks, and paradigms, and that comprehenders under certain circumstances try to get rid of the negation by transforming the negative sentence into a meaning-equivalent affirmative sentence. This early research has laid the foundation stone for many of the research questions in the following years, which we will discuss in the following sections. This holds for instance for the assumption that negation is an explicitly represented operator that in a propositional representation encapsulates the negated information. This assumption sparked the hypothesis that one important function of negation is to reduce the availability of information in its scope and thus routinely involves suppression mechanisms. This research will be discussed in section .. In addition, with the development of alternative theories regarding the representational format of meaning representations created during language comprehension, the question came up whether the meaning of negation can also be represented in a nonpropositional format. Here research focused on the hypothesis that negation is implicitly represented in the sequence of representations that a sentence gives rise to. We will discuss this research in section ., which deals with the question whether negation processing routinely involves representing two states of affairs, the non-factual and the factual state of affairs. A further research question that connects later negation research with the already reported studies on the processing of negation concerns pragmatic aspects of negation. Research here focuses on the hypothesis that negation integration during comprehension is typically delayed and requires special attention but that it may be facilitated and proceed in a more automatic fashion when negative sentences are presented in a pragmatically felicitous context. We will discuss this research in section .. Finally, research in later years also focused on novel questions that were not already part of the early research. One important question is how negation influences embodied representations during language comprehension. Another important question that has only recently received attention in the literature concerns the idea that negation processing may involve general cognitive mechanisms that are well known from research on non-linguistic cognition such as the processes that are involved when a prepared action needs to be cancelled or conflict-related adaptation processes. We will deal with this more recent research in sections . and ..

.. I      ?

.................................................................................................................................. In the research reported in section . researchers assumed that the meaning representation of a negative sentence involves an explicit representation of the negation operator as well as the proposition representing the non-factual states of affairs (e.g. Sam does not wear a hat is represented by ‘not[wear[Sam,hat]]’ with Sam wearing a hat being the non-factual state of affairs). MacDonald and Just () were the first to experimentally investigate whether the negation operator would reduce the activation of information represented in its scope. Their hypothesis was based on the intuition that concepts introduced within the

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 



scope of a negation operator are less relevant for the ongoing discourse than concepts introduced in affirmative phrases. This in turn also fits well with the results of linguistic analyses according to which discourse referents introduced in negated phrases are not readily available for subsequent pronominal reference (e.g. Kamp and Reyle : ). To test their hypothesis, MacDonald and Just presented participants with sentences such as Every weekend Mary bakes bread but no cookies for the children and, immediately afterwards, measured the accessibility of the relevant concepts by means of a probe recognition or word naming task. Longer response times were observed for negated compared to nonnegated probe words (cookies vs. bread, respectively), independent of the position of the negation in the sentence. This result seems in line with the idea that readers construct a propositional representation in which the negation operator encapsulates the negated information and thereby reduces the availability of this information. Findings by Moxey and colleagues on negative quantifiers (Moxey and Sanford ; Paterson, Sanford, Moxey, and Dawydiak ; Sanford, Moxey, and Paterson ) can also be interpreted as evidence for the view that negation functions as an activation-reducing operator: Sentences beginning with negative quantifiers (e.g. Few of the football fans went to the game) differ from the corresponding sentences with a positive quantifier (e.g. Many of the . . . ) in how a subsequently encountered plural anaphor (e.g. they) is interpreted. For positive quantifiers, it is interpreted as referring to the reference set of the sentence (i.e. the subset of fans who went to the game), whereas for negative quantifiers, the bias is toward the complement set (i.e. the subset of fans who did not go to the game). In other words, a negative quantifier seems to shift the discourse focus away from the subset explicitly referred to in the sentence, which is similar to the idea that negation reduces the activation of the information mentioned in its scope. A very different position is occupied by Giora and her colleagues (Giora et al. ; Giora et al. ; Giora, Zimmerman, and Fein ). She argues that suppression is not the default function of the negation operator but may be a result of the fact that the sentences in the studies by MacDonald and Just () and Moxey and colleagues were presented in isolation. According to Giora, information presented in longer discourses is retained or discarded as a function of more general discourse cues indicating a continuation or shift in topic (cf. Gernsbacher ). In line with this assumption, a target concept such as fast was primed rather than suppressed after a sentence such as The train to Boston was no rocket when followed by a coherent sentence that continued rather than shifted the topic of the earlier sentence (e.g. The trip to the city was fast though as compared to The old man in the film spoke fast; Giora et al. ). Similarly, after reading a discourse such as I live in the neighborhood of millionaires who like only their own kind. Nonetheless on Saturday night, I also invited to the party at my place a woman who is not wealthy the related probe rich was primed rather than suppressed. The priming of the related probe was not due to the context sentence itself, as no priming was observed for the same context sentence that ended in an affirmative phrase (e.g. and my sister lives in Haifa in a neighborhood that is religious). Both results indicate that suppression is not always involved in the processing of negation, at least not early on during the respective comprehension processes. There are other studies that also suggest that negated probes are not always reduced in activation after the processing of negation (cf. Autry and Levine , ; Levine and Hagaman ). However, as the authors of these studies explain their results in

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



    

terms of pragmatic processing, we will not discuss them in detail here but rather in section ., which deals with pragmatic aspects of negation processing. As mentioned, Giora attributes results such as those observed by MacDonald and Just () to the fact that sentences were presented in isolation. An alternative explanation suggests itself from the perspective of the situation-model or mental model theory (van Dijk and Kintsch ; Johnson-Laird ), according to which comprehenders do not only construct text-based propositional representations but also non-linguistic representations of the state of affairs that the text is about. These so-called mental models or situation models are characterized as being structurally analogous to the states of affairs they represent. In particular they are assumed to contain tokens that stand for the entities that are part of the described situation as well as to represent properties and relationships that are true for these entities. Thus, for example, the model built for the sentence Sam did not wear a hat should not contain a token for a hat, since there is no hat in the situation described. Assuming that such a model is tapped in word-recognition and word-naming tasks, the result of MacDonald and Just can easily be explained: In the situations described in their sentences the non-negated probe words name entities that are present (cookies) whereas the negated probe words do not (bread). In line with the idea that the presence in the described situation may be an important variable, texts such as Elizabeth tidies out her drawers. She burns the old letters but not the photographs. Afterwards she cleans up in which the negated entity is present in the described situation do not lead to a reduced availability of the negated concept (Kaup , ). In this case, however, the availability of the relevant concepts was measured with quite a delay after processing the negated sentences. Maybe activation-reducing effects of negation can only be observed at an earlier point in time during the comprehension process when comprehenders are still engaged in creating the linguistic representation of the text (the so-called text base, cf. van Dijk and Kintsch ; see also Kaup and Zwaan  for a study directly investigating this issue). We would like to point out that there are also a number of studies that more directly indicate that negation may involve suppression processes. As these studies were conducted in the context of the embodied cognition framework and investigated the question how embodied effects in language comprehension are affected by the presence of a negation operator in the sentences, we will discuss these studies in more detail in section ., which deals with negation and embodiment. For the present purpose, it suffices to say that it is difficult to see how these results can be interpreted without assuming that negation processing somehow involves inhibitory mechanisms, at least as far as action-related language is concerned. A further indication that negation processing may involve inhibition is provided by a study by Mayo, Schul, and Rosenthal () showing that successfully negating a false predicate of an entity (Was the dress blue? No) reduced the availability of the entity (i.e. the dress) in a later memory task when compared to positively responding to a question concerning a true property (Was the dress red? Yes). The authors interpret this finding as evidence for inhibitory processes that over-generalize from the property to the whole entity. We will come back to this issue in section ., which deals with the question whether negation processing involves general cognitive mechanisms known from research on non-linguistic cognition. To summarize, the question whether negation processing routinely involves suppression has been investigated in numerous studies with quite different methodologies and paradigms. Although the results of the different studies are not completely consistent, it seems

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 



that there is good evidence that negation processing sometimes involves some kind of suppression mechanism by which the information presented in its scope is reduced in availability. The temporal characteristics of this suppression mechanism are less clear. Whereas some authors assume that suppression happens early on during the comprehension process, others suggest that it takes place at a later point in time. The interpretation of the results of studies concerned with this issue is hampered by the fact that we do not know whether there is necessarily a fixed order with which the involved processes take place. In principle, it seems possible that the order also depends on the complexity of the involved subprocesses (cf. Autry and Levine ). In any case, as was mentioned above, we think it is safe to conclude that suppression processes are part of negation processing even though the exact nature and the temporal characteristics of these mechanisms are unclear.

.. D                  ?

.................................................................................................................................. One question that continues to play a central role in the literature on negation processing concerns the question whether comprehenders process negation in one or two processing steps, and accordingly involves one or two representations. According to two-step accounts (e.g. Fauconnier ; Kaup et al. ; Langacker ), the processing of a negative sentence (e.g. Sam does not wear a hat) involves two steps during which comprehenders first represent the non-factual state of affairs (Sam with a hat) and then the factual state of affairs (Sam without a hat). Hereby, negation is implicitly captured in the deviations between the two representations. According to a one-step account (e.g. the fusion model suggested by Mayo et al. ), the factual state of affairs would be represented right away. Evidence for two-step-accounts comes from studies in which the availability of the two different states of affairs was measured with different delays after reading negative sentences. The reasoning was that shortly after reading a negative sentence, comprehenders might still be engaged in the first processing step and accordingly have available a representation of the non-factual state of affairs. In contrast, after a certain delay they may have arrived at the second step and thus have available a representation of the actual state of affairs. In line with this prediction, readers of negative sentences responded faster to a picture that matched the non-factual rather than the factual state of affairs when the picture was presented with a short delay of  ms after the sentence, but faster to a picture that matched the factual rather than the non-factual state of affairs when the picture was presented with a long delay of  ms after the sentence (Kaup, Lüdtke, and Zwaan ; see also Anderson et al. ; Kaup, Lüdtke, and Zwaan ; Kaup et al. ; Lüdtke et al. ; Scappini et al. ). Similar results were observed when the availability of the respective representations was measured by means of a lexical decision task. For instance, in a study by Hasson and Glucksberg (), participants read affirmative and negative metaphors (e.g. Some surgeons [are] / [aren’t] butchers) and afterwards responded in a

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



    

lexical decision task to words that were semantically associated with the factual or with the non-factual state of affairs implied by the metaphors (e.g. clumsy vs. precise, respectively). As predicted, when tested with short delays after sentence processing (i.e.  and  ms), negative metaphors facilitated responses to associations of the non-factual state of affairs, whereas with a  ms delay, facilitation could no longer be observed. In contrast, the affirmative metaphors facilitated responses to associations of the factual state of affairs across all three delays. As the target words were not lexically associated to any of the words in the sentences, this result indeed suggests that comprehenders in this case initially considered the non-factual state of affairs during the comprehension of the negative metaphors. However, this result of course does not indicate that representing the nonfactual state of affairs is a necessary intermediate comprehension step in the processing of negative sentences. Many authors have suggested that comprehenders represent the nonfactual state of affairs only under special circumstances. Two conditions in particular were intensively discussed in the literature. First, it has been suggested that the non-factual state of affairs is only represented if the implied Question under Discussion (QUD, cf. Roberts ) is affirmative rather than negative, and the results of a number of studies are in line with this prediction. For instance, participants in a visual world study by Tian and Breheny () indeed showed different eye-movement patterns when reading a negative sentence with an affirmative versus a negative QUD (e.g. John didn’t iron his brother’s shirt—QUD: Did John iron his brother’s shirt? Vs. It was John who didn’t iron his brother’s shirt—QUD: Who didn’t iron his brother’s shirt?). With affirmative QUDs, participants early on focused on the non-factual state of affairs (i.e. an ironed shirt) whereas with negative QUDs, participants immediately focused on the factual state of affairs (i.e. a wrinkled shirt; see also Tian, Breheny, and Ferguson ). These results also fit nicely with the idea that one reason the non-factual state of affairs often gets activated during the processing of negative sentences has to do with the pragmatics of negation (see above): Negative sentences with an affirmative QUD correct the assumption that the non-factual state of affairs is true, whereas negative sentences with a negative QUD do not. Second, many authors assume that the non-factual state of affairs is only represented when the negative sentences refer to an attribute dimension for which there are only two as opposed to multiple values in the reference situation (reading ‘not black’ in a situation where an object is either black or white as compared to a situation in which the object can be of many different colors). We will refer to these two types of negation as ‘binary’ vs. ‘non-binary’, respectively. The availability of the non-factual state of affairs after processing negative sentences has indeed been shown to be influenced by the difference between these two types of negation. Evidence stems from studies using different paradigms and dependent variables (Du et al. ; Mayo, Schul, and Burnstein ; Orenes, Beltrán, and Santamaría ). Before turning to the next question, we would like to emphasize that the literature on ironic effects of negation also is in line with the idea that comprehenders under certain conditions involuntarily activate the non-factual state of affairs when processing negation, even if this is counterproductive to an effective mastery of the task. For instance, in a study by Wegner, Ansfield, and Pilloff () participants were more likely to overshoot a golf ball when explicitly told to avoid this by means of a negative instruction (e.g. Land the ball

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 



on the glow spot vs. . . . don’t overshoot the glow spot), especially when put under mental or physical load. Ironic effects have been reported in many other domains as well (e.g. thinking about anxiety: Koster et al. ; thinking about sex: Wegner et al. ; prejudice: Gawronski et al. ; but see Johnson, Kopp, and Petty ). Although not explicitly discussed in this literature, many of the implemented negative instructions indeed involved non-binary rather than binary negative sentences. These results therefore are compatible with the idea that the non-factual state of affairs is predominantly activated for non-binary as opposed to binary negative sentences. In fact, in the literature on intention implementation, forming a ‘negation implementation intention’ (e.g. When I am sad, I will not eat chocolate) resulted in a strengthening rather than weakening of the habit compared to a replacement implementation intention (When I am sad, instead of eating chocolate I will . . . ; Adriaanse et al. ) suggesting that the activation of the non-factual state of affairs is dependent on how available the alternative state of affairs is to the comprehenders. To summarize, there is evidence that comprehenders of negative sentences mentally represent the non-factual state of affairs in addition to the factual state of affairs. However, some authors have argued that this is only the case in special circumstances, for instance when the sentence involves a non-binary negation and the comprehender therefore does not have much information regarding the factual state of affairs. The omnipresence of ironic effects of negation also seems to suggest that negation triggers activating the nonfactual state of affairs under certain conditions. However, in most of the cases, these ironic effects could also be word-based and come about because comprehenders do not succeed in immediately integrating the negation operator with the meaning of the rest of the sentence. Section . will deal with studies that investigated the question under which circumstances negation integration is delayed during comprehension.

.. I                   ?

.................................................................................................................................. As was already mentioned, negation often seems to lead to comprehension difficulties resulting in rather long processing times and/or incomplete meaning representations shortly after the processing of negative sentences. An obvious question that suggests itself in the context of such findings is to what extent negation processing is automatic. The question here is not whether the negation operator itself is processed automatically. Indeed, there are studies suggesting that comprehenders routinely and quickly register whether a sentence is negative or not, as indicated for instance by the fact that negative polarity items (e.g. any, ever) elicit smaller N amplitudes in licensed contexts involving a negation marker than in unlicensed contexts without a negation marker (Xiang, Grove, and Giannakidou ). The question is rather whether the meaning change that the negation marker indicates is determined automatically.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



    

... Automaticity The studies discussed in section . (increased processing times for negative sentences) already seem to question the idea that negation processing might proceed automatically. Indeed, some subsequent research directly addressing this question suggests that negation integration requires controlled processes. In a study by Deutsch, Gawronski, and Strack () participants evaluated phrases involving words with positive or negative valence (e.g. party vs. war, respectively), either in an affirmed or in a negated version (e.g. a party vs. no party; a war vs. no war, respectively). Although overall response times in this task clearly decreased in the course of the experiment, the difference between response times for affirmative and negative trials remained constant. This pattern of results therefore seemed to suggest that accessing the valence of the nouns in the phrases was facilitated by practice, but the valence reversal mechanism required by the negation was not. Thus, it seems that negation integration cannot be automatized by extensive practice. In addition, in several evaluative priming experiments, the authors demonstrated that nouns with a positive valence facilitated the processing of subsequent positive targets, independent of whether the primes appeared in an affirmative or negative phrase ([a]/[no] friend primed luck), and the same was true for nouns with negative valence ([a]/[no] war primed disease). Again, this suggests that negation integration is not automatic, this time in the sense that it does not occur unintentionally in a task that does not require it. However, later research indicated that under certain conditions, negation integration does seem to occur fast and unintentionally in a task that does not require it. Deutsch et al. () presented their participants with pairs of words consisting of a noun and an affirmative or negative qualifier (e.g. no sunshine), followed by Chinese ideographs and asked them to judge the visual pleasantness of the Chinese ideographs as either above or below average. The authors observed an interaction between the valence of the noun and the polarity of the phrase suggesting that the valence of the priming phrase was affected by the negation, and thus indicating that comprehenders had integrated the negation and the noun even though the task did not require this. Thus, negation integration had taken place quickly and unintentionally. A similar conclusion was also drawn by Armstrong and Dienes () based on observing priming effects of subliminally processed word pairs involving a negation marker. However, in the study by Deutsch et al. () an interaction between valence and polarity was not observed when participants performed the task while being distracted by a digit memory task, showing that negation integration may proceed quickly and unintentionally but not independent of cognitive resources (see also Herbert and Kübler ). Maybe this is also the reason why in sentence comprehension, which is a resourcedemanding process in itself, negation integration often does not seem to be completed when comprehenders reach the end of the sentences. Maybe, there are simply not enough resources left during normal sentence comprehension to rapidly integrate the negation into the sentence meaning, leading to a delayed integration. This fits well with the results of the many sentence comprehension studies that observed the same response time / N results whether the sentence contained a negation marker or not. To illustrate, sentences such as Ladybirds are (not) stripy lead to longer response times and higher N amplitudes on the final word compared to sentences such as Zebras are (not) stripy, independent of whether

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 



there is a negation in the sentence (Giora et al. ; Hald et al. ; Kounios and Holcomb ; see also Dudschig and Kaup ; Dudschig et al. ; Dudschig et al. ). Many authors have pointed out, however, that the conclusions that negation integration is generally delayed in sentence comprehension may be premature. In many of the studies suggesting a delay in negation integration, negative sentences were presented outside of a context that would pragmatically license the negation. As was mentioned above, in real discourse, negative sentences are usually used to indicate deviations from expectancies or to correct a false proposition (cf. Givon , ). When not used in this way negative sentences are pragmatically infelicitous. In this case, comprehenders possibly accommodate the implied expectancies or false propositions as part of comprehending the negative sentence and this may in turn delay negation integration. If so, negation integration can be expected to proceed in a timelier manner in case negative sentences are used in a pragmatically felicitous way. We will discuss the studies that investigated this prediction in the next paragraph.

... Pragmatically felicitous contexts Early studies investigated the prediction that negative sentences are easier to process when they occur in a pragmatically licensing context. The underlying assumption was that licensing is particularly important for negation, more so than for affirmation. For instance, in a study by Glenberg et al. () participants were presented with longer narratives in which the penultimate sentence was either negative or affirmative (e.g. The couch [was] / [was not] black). In the supportive-context condition, the previous text highlighted the relevancy of the attribute dimension that was referred to in the penultimate sentence (e.g. She wasn’t sure if a darkly colored couch would look the best or a lighter color). In contrast, in the non-supportive context condition, the prior text highlighted a different attribute dimension (e.g. She wasn’t sure what kind of material she wanted the couch to be made of). As predicted, negative target sentences took longer to process than affirmative ones, but only in the non-supportive contexts. Thus, the processing of negative sentences was facilitated by a supportive context, more so than the processing of the affirmative sentences was. Similarly, in a study by Lüdtke and Kaup () the processing of negation was facilitated when it was presented in contexts in which the proposition that was being negated was either explicitly mentioned as a potential possibility in the prior text (e.g. . . . She wondered whether the water would be warm or cold . . . The water was not warm) or at least constituted a highly plausible assumption in the context (Betty’s young son was not shy and participated in any nonsense that the kids could come up with . . . his T-shirt was not dirty). Interestingly, a facilitative effect of such contexts was not observed for individuals diagnosed with Asperger’s syndrome who presumably engage less in pragmatic language processing compared to healthy individuals (Schindele, Lüdtke, and Kaup ). Together, these studies suggest that the processing difficulties associated with negation are (at least partly) due to issues related to pragmatic licensing. Nordmeyer and Frank (unpublished manuscript) draw a similar conclusion based on a study that directly compared speakers’ and listeners’ production and comprehension of affirmative and negative sentences in different contexts. For instance, when seeing a picture of a particular boy (say Bob) in

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



    

the context of four other boys who either all do or do not carry an apple, then speakers who aim at describing Bob are more likely to produce a negative sentence (e.g. Bob has no apple) in the former than in the latter case, and this is also reflected in the processing difficulties that comprehenders have with negative sentences in these two contexts. This finding is nicely in line with the idea that negative sentences are difficult to process mainly in contexts in which they are unlikely to be produced, suggesting that negation is not particularly difficult as long as it is used in pragmatically licensing contexts. Finally, there is evidence that in certain contexts, negative terms (e.g. impossible) may even be easier to integrate than the corresponding affirmative term (e.g. possible). When participants read these expressions in an adequate context (e.g. Creating such a machine is very hard, but it is [possible] / [not impossible]) the negative term elicited an N with a smaller amplitude than the affirmative term, suggesting that negation in some contexts can function as a cohesion marker (Schiller et al. ). In addition to these studies that looked at differences in processing time required for negative sentences appearing with or without a licensing context, there are a number of studies that investigated in more detail how different aspects of negation processing are influenced by contextual information. We will discuss two lines of research. The first line of research investigates the hypothesis that negation integration is delayed when negative sentences violate pragmatic requirements but can take place during online comprehension for pragmatically felicitous negative sentences. The results of a study by Nieuwland and Kuperberg () are in line with this prediction: N amplitudes for target words (e.g. safe/dangerous) showed a truth value by polarity interaction for pragmatically infelicitous sentences (Scuba-diving is not safe), a pattern that suggests that comprehenders had not yet successfully integrated the negation operator with the meaning of the rest of the sentence when they encountered the target word. In contrast, pragmatically felicitous negative sentences (e.g. With proper equipment, scuba-diving is not dangerous), showed a main effect of truth value, a pattern that indicates successful negation integration. In this study, pragmatic licensing was independently assessed in a naturalness-rating prior to the experiment, distinguishing between true-negated sentences that were relatively informative (With proper equipment, scuba-diving isn’t very dangerous) and true negative sentences that were under-informative or trivial (Scuba diving isn’t very safe). One could object, however, that in this experiment, pragmatic licensing may have been confounded with predictability. Indeed, in a later follow up study, Nieuwland () showed that negation integration was facilitated in sentences in which the target words were relatively predictable as measured by an independent cloze-test prior to the experiment. Whether predictability of the relevant words or pragmatic licensing is the more relevant factor when it comes to negation integration cannot be determined on the basis of the available evidence. There are also studies that investigated whether pragmatically felicitous contexts would change the number of processing steps that are required by a negative sentence (cf. the discussion in section . above). In a study by Dale and Duran (), participants verified simple affirmative and negative statements (e.g. Elephants are [large]/[not small]) by means of moving a mouse towards a response field. In line with the idea that negation processing proceeds in two steps, negation caused more discreteness in the mouse trajectory of a response. When the negative sentences were embedded in pragmatically felicitous contexts (You want to lift an elephant? But elephants . . . ), however, discreteness in the mouse

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 



trajectories was significantly diminished, suggesting that negation processing now involved only one processing step. The second line of research explicitly investigated the assumption that negation —when presented in infelicitous contexts—leads to a reactivation of the non-factual state of affairs as a consequence of accommodating the presuppositions that a negative sentence carries, but that this step is unnecessary when negation is presented in felicitous contexts. In line with this prediction, negated concepts proved to be rather highly accessible after reading pragmatically infelicitous negative sentences. In this case, negated concepts, for instance, interfered with anaphor resolution: He ate the fruit in the kitchen was more difficult to process after reading Justin bought a mango but not an apple than after Justin bought a mango but not any water suggesting that the apple was considered as a potential antecedent (Levine and Hagaman ). In addition, negated concepts often occurred in the continuations that comprehenders produced after reading the negated sentences (Autry and Levine ). Finally, processing times for words that had been mentioned in negated phrases were relatively low when availability was tested with a certain delay after the processing of the sentences. In contrast, after reading negative sentences in a pragmatically felicitous context, processing times for probe words that were mentioned in negated phrases were relatively high, presumably indicating that in this case no reactivation of the non-factual state of affairs had taken place (Autry and Levine ). These results are in line with the view that negation processing leads to a representation of the non-factual state of affairs in particular in contexts in which this state of affairs was neither explicitly mentioned in the prior discourse nor constitutes a particularly plausible assumption (cf. Kaup ). Also in line with this idea are the results of studies employing a visual-world paradigm while participants are reading affirmative and negative sentences in different contexts. In a study by Orenes et al. (), for instance, participants more quickly preferred the factual over the non-factual state of affairs after reading a negative sentence (e.g. her dad is not poor) in a pragmatically felicitous context (e.g. She supposed that her dad had little savings) compared to a pragmatically infelicitous or neutral context (e.g. She supposed that her dad had enough savings and Her dad lived at the other side of town, respectively) suggesting that the pragmatically felicitous context facilitated negation processing by preactivating the non-factual state of affairs. Finally, the above-mentioned visual-world studies by Breheny and colleagues showing that the non-factual state of affairs is fixated early on during the processing of negation only when the implied QUD is affirmative also fit well into the current debate, as negative sentences with affirmative QUDs correct a false assumption that presumably has to be accommodated during the processing of negation when used in a pragmatically infelicitous sentence (Tian and Breheny ; Tian, Breheny, and Ferguson ). Taken together, the results reported in this section show that negation processing is facilitated when a negative sentence occurs in a pragmatically felicitous context. It seems that in this case, comprehenders are faster at integrating the negation operator with the meaning of the rest of the sentence such that they have available a sentence-based meaning representation when they reach the end of the sentence. The studies also indicated that the speed up in negation processing in licensing contexts is likely due to the fact that comprehenders in these contexts do not need to accommodate presuppositions and activate the non-factual state of affairs in the course of negation processing.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



    

. N  

.................................................................................................................................. The concept of embodiment has received a lot of attention in recent years in language research. However, experimental studies conducted in this conceptual framework predominantly looked at the comprehension of affirmative sentences. However, how negation can be captured in embodied representations is an important issue. The present section deals with this question, starting with a short introduction to embodied accounts of language comprehension in general. According to an embodied account (Barsalou ; Glenberg and Gallese ) language is directly interrelated with the sensorimotor system, resulting in modal rather than abstract-symbolic representation of linguistic meaning. These embodied models of language comprehension typically follow an explanatory approach that is strongly linked to assumptions regarding language learning (see also Vogt, Kaup, and Dudschig ). When a child learns language this typically occurs in a multimodal environment, whereby linguistic and non-linguistic experiences co-occur. For example, if a child encounters the word lemon this most likely will be in situations where an older sibling handles a lemon or even passes a lemon slice for tasting. The child will therefore at the same time hear the word lemon and experience various sensory aspects of a lemon (e.g. color, taste, shape). Strong embodiment models propose that when the word lemon is later encountered in another setting, these earlier sensorimotor experiences become reactivated and build the core of understanding. Indeed, there exists a rather large amount of evidence supporting such assumptions for single word and sentence comprehension (e.g. Glenberg and Kaschak ). However, can these models also explain the comprehension of constructions with linguistic operators such as negation that do not directly have a to-be-experienced counterpart in the real world? Whereas for other types of abstract language—for example the information about time—there have been manifold suggestions how understanding can still be embodied, for example via a metaphoric mapping account (e.g. Ulrich and Maienborn ), this question is still debated with regard to negation. The first studies dealing with negation in the context of the embodied cognition account investigated the hypothesis that negation is implicitly represented in the sequence of mental simulations that a sentence gives rise to. These studies were already discussed in section ., but at that point without referring to the underlying assumptions concerning representational format. The more specific assumption that these studies investigated was that comprehenders of negative sentences in a first step mentally simulate the non-factual state of affairs, then turn their attention away from this simulation, and instead engage in a simulation of the factual state of affairs (e.g. Kaup, Zwaan, and Lüdtke ). To investigate these specific predictions, the above-mentioned studies built upon effects that for affirmative sentences are typically interpreted as evidence that comprehenders activated mental simulations during comprehension. For illustrative purposes, consider the finding that comprehenders are faster to identify the picture of an eagle if depicted with outstretched compared to folded wings after reading The ranger saw the eagle in the sky, whereas the opposite is true after reading The ranger saw the eagle in the nest (Zwaan, Stanfield, and Yaxley ). In the respective studies dealing with negation, the same phenomenon was now investigated but by manipulating both the polarity of the sentences (affirmative vs.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 



negative) as well as the delay with which the picture was presented after reading the sentences. Across a number of studies, this approach showed that the investigated embodied effects varied with the interaction of polarity and delay: Differences between affirmative and negative sentences occurred only after certain delays after comprehending the sentences. These results were interpreted as positive evidence for the view that comprehenders first simulate the non-factual state of affairs followed by simulating the factual state of affairs during the processing of negative sentences (for details see Kaup, Zwaan, and Lüdtke  and the studies reported therein). Later studies more directly investigated the question how embodied effects during language comprehension are affected by the presence of a negation operator in the sentences. Most of these later studies specifically investigated the hypothesis that negation would suppress embodied effects or hinder them from occurring, for instance by triggering inhibitory processes. In a study conducted in our lab, participants read affirmative sentences describing an up- versus down-movement (The lion is climbing upwards/downwards) as well as their corresponding negative counterparts (The lion is not . . . ). In a sensibility-judgment task, participants then responded with either an upwards directed or downwards directed response movement. Affirmative sentences showed a compatibility effect between the spatial term in the sentences (upwards vs. downwards) and the response movement, whereas negative sentences did not show such a compatibility effect (Dudschig, de la Vega, and Kaup ). Similarly, an fMRI study by Tettamanti et al. () aimed to directly measure action-related motor activation during negation integration. In that study participants read Italian action-related sentences (e.g. Adesso io rem oil bottone [Now I push the button], Adesso non rem oil bottone [Now not push the button]) versus abstract sentences (Ora io apprezzo la fedeltà [Now I appreciate the loyalty], Ora non apprezzo la fedeltà [Now not appreciate the loyalty]). In line with the embodied language processing model, action-related sentences resulted in stronger activation of the action-representation network in contrast to abstract sentences. For both abstract and action-related sentences, negations resulted in a deactivation in the pallido-cortical regions. Interestingly, negation significantly reduced the involvement of the action-representation network, specifically when processing action-related sentences. The authors concluded that negation reduces access to the embodied representation of the affirmative counterpart. Tomasino, Weiss, and Fink () replicated these findings in a similar experimental setup. They used short imperative statements not containing an object noun (e.g. Don’t grasp, Do write) and showed in line with Tettamanti et al. () that sentence polarity influences the activation of motor circuits in the brain. Specifically, negation again had an inhibitory effect on the activation of motor-related conceptual networks. In a follow-up study by Bartoli et al. () it was investigated whether the access reduction in embodied representational networks frees resources for parallel tasks during negation integration. The authors used affirmative and negated sentences that were either abstract in nature (e.g. I do not wish), sentences describing movements involving distal muscle movement (hand/finger muscles: e.g. I do not paint) and sentences describing proximal muscle movements (shoulder/ arm muscles: e.g. I do not post). After sentence presentation, participants had to perform either proximal arm movements (reach-to-grasp), or distal arm movements (grasping movement). The results showed that negated sentences in both the proximal or distal domain eased the performance in the concurrent arm movement task. It was argued that in line

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



    

with the previous study (Tettamanti et al. ) negation seemed to result in reduced activation of specific embodied representational networks (and therefore frees these resources for performing the according arm movements). These findings were further supported in a study by Aravena et al. () using a novel grip-force measurement apparatus with participants constantly holding the apparatus during auditory encountering of an affirmative or negated sentence (e.g. . . . lifts the dumbbells, . . . doesn’t lift her luggage, . . . loves the flowers). Online measurements of grip-force showed that lexical semantic processing of the action verbs increased the grip-force but only in the affirmative sentences. In the negated sentences implementing action-verbs and the sentences describing non-action related information, the grip-force stayed at baseline level. In line with the previously described fMRI studies the authors concluded that negation results in inhibition of motor-related activation. All these studies seem to suggest that negation affects embodied effects by triggering inhibitory mechanisms that suppress or hinder these embodied representations from occurring. On the one hand, this result seems in line with the embodied cognition account. When reading about someone not grasping a pen, then the described state of affairs does not involve a grasping movement and accordingly the motor system should not be activated. However, on the other hand, one could also argue that the meaning of the negated sentence then is not fully captured in an embodied representation created during comprehension. After all, how can one in this case, on the basis of the created embodied representations, distinguish between the information that someone did not grasp a pen and the information that someone did not pick an apple, or even worse, the information that someone did not think of his girlfriend? It seems that the embodied representations are only consistent with the linguistic information given but are not specific enough to capture the full meaning. As a consequence, some authors have taken results of this type as a basis for proposing a disembodied account of negation (e.g. Bartoli et al. ; Tettamanti et al. ). There is one study however that speaks against a disembodied account of negation: Foroni and Semin () investigated the processing of negated sentences that directly refer to a specific facial expression. Participants read sentences describing facial movements that involve the zygomatic muscle (e.g. I am smiling) or do not involve this muscle (e.g. I am frowning) in either affirmative or negated versions. Interestingly, the results showed that negating a statement that in its affirmative version would involve a zygomatic muscle activation, resulted in a below-baseline inhibitory activation of this particular muscle. In this case, one could argue that negation is captured in an embodied manner during comprehension. In summary, various studies addressed the question how embodied effects are influenced by a negation marker in the sentence. Most of these studies showed that negation reduces or prevents these effects from occurring. These studies thus provide evidence that embodied representations are sensitive to the negation in the sentences but might not fully capture the meaning of negative sentences. On the other hand, there is first evidence that at least in some cases, negation may lead to more specific embodied representations. One problem when interpreting these results may be given by the fact that many of these results may only reflect the end state of comprehension. As mentioned above, negation could in principle also be represented implicitly in the sequence of meaning representations that get activated during comprehension. A full understanding of the embodied representation of negation thus requires additional research that looks into more detail concerning the temporal dynamics of embodied representations activated by negated sentences. Interestingly,

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 



recently authors have attempted to address this question from a modeling approach. Huette and Anderson () developed a recurrent network with a grounded perceptual simulation layer that incorporates negation. This network was successfully trained and tested on new material, and thus points towards the option that negation can indeed be comprehended without the need for any abstract/symbolic processor. Nevertheless, future studies will be needed to get a comprehensive idea of the involvement of embodied and abstract processes in negation integration.

.. D            -     ?

.................................................................................................................................. In recent years, a new question regarding negation integration processes has emerged. Motivated by the earlier described suppression and inhibitory mechanisms that are often reported during negation integration, and at the same time the embodiment account, claiming a strong interrelation between linguistic and non-linguistic cognition, the question arose, whether negation processing uses or falls back on mechanisms that are also utilized in non-linguistic cognitive processes. Specifically, three processes were under the focus of investigation: inhibitory mechanisms known from memory processes, inhibitory processes utilized in action processes, and cognitive control mechanisms typically reported during action control. One of the first studies that investigated the interplay between negation and general cognitive mechanisms was a study investigating whether negation results in forgetting (Mayo, Schul, and Rosenthal ). In the memory literature it is established that the instruction to ‘not-think’ about something results in a below-baseline memory rate for these no-think items (Anderson and Green ). Mayo et al. () investigated whether simply using a negation results in forgetting these items. In this study, participants saw for example a picture of someone drinking white wine. Afterwards they were asked Was it [red]/[white] wine? and correctly answered ‘no’ or ‘yes’, respectively. Interestingly, at the end of the experiment participants showed a greater memory loss for the wine-drinking picture when they had been in the ‘no’ condition than when they had been in the ‘yes’ condition. However, overall the memory rate in the negation condition was still larger than for items that were not followed by a question. Thus, being asked a question about a picture does increase memory, however this increase is smaller in the ‘no’-answer condition. These findings point towards the idea that negation use relies on cognitive mechanisms that also influence memory processes. In an electrophysiological study, De Vega et al. () investigated whether inhibitory mechanisms that are well studied in the action control literature are activated by negated sentences. Participants read affirmative and negated sentences (e.g. Now you [will]/[will

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



    

not] cut the bread). Shortly after verb onset ( ms), a visual cue (circle) was presented above the verb that indicated whether participants should press a button or withhold their responses. The results showed that theta oscillation—a neural marker of response inhibition—was significantly reduced for NoGo trials if they occurred in negative sentences, supporting the idea that negation processing recruits similar neural mechanisms as are active during action control (for similar results with non-action-related sentences see also Beltrán et al. ). In a follow-up study by Beltrán, Muñetón-Ayala, and de Vega (), participants read affirmative and negative sentences (e.g. You [will]/[will not] cut the bread). A right or left pointing arrow was displayed above the verb, and participants were asked to press the respective response key. In some trials, an auditory Stop-signal was presented, indicating to the participant to withhold their response. Electrophysiological results showed that inhibition-related N/P components were enhanced in successful inhibition trials. The N (source-located in the right inferior frontal gyrus) was also modulated by sentence polarity, being largest in negative successful inhibition trials. Additionally, the estimates of stop-signal reaction times in negative trials were increased, pointing towards an interaction of negation processing and inhibitory control processes used for action control. In a very recent study by García-Marco et al. () it was shown that negation comprehension also has an inhibitory influence on standard typing responses. Participants read two successive sentences (e.g. There is a contract. Now you [are]/[are not] going to sign it) and then typed the final verb on a keyboard. Participants were significantly slower to type manual-action verbs if these occurred in a negated phrase. Thus, this study again indicates that negation has inhibitory influences on non-linguistic typing responses, which points towards the idea that inhibitory mechanisms are shared between the linguistic and the non-linguistic system. In a final line of research—addressing the overlap between linguistic and action-related processes—a study conducted in our lab (Dudschig and Kaup ) investigated whether negation results in conflict-like processing adjustments. In the action control literature it is well established that conflict during information processing (e.g. an incongruent Stroop, Flanker, or Simon trial) results in subsequent processing adjustments (e.g. Botvinick et al. ). These adjustments result in easing the processing of conflicting trials that follow another conflicting trial. In our study conflict was induced by means of negation; participants had to respond to the phrases not left, now left, not right, and now right with according button presses. Indeed, in our study processing adjustments were reported that resembled those in nonlinguistic conflict tasks. Despite being particularly slow for negated trials, this slow-down was reduced if the negation trial was preceded by another negation. Overall, this rather new line of research investigating the interaction between negation processing and non-linguistic inhibitory or control processes points towards direct overlaps in the way inhibitory and control processes work in linguistic and non-linguistic information processing.

.. C

.................................................................................................................................. The research reported in this chapter has shown that processing negation is often a timeand resource-consuming mental process. However, there are also circumstances where

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 



negation comprehension seems to proceed without difficulty. There are different reasons why negation processing is often difficult. One reason may be that comprehenders of negative sentences in many cases need to create two meaning representations, one of the non-factual state of affairs and one of the factual state of affairs. Research however has also shown that creating two meaning representations is not required under all circumstances. For instance, when negative sentences are presented in a pragmatically licensing context, the non-factual state of affairs is often pre-activated or easy to retrieve, and representational efforts are therefore reduced. Indeed, in these contexts, negative sentences are not only processed much faster, negation integration also has a higher chance of being completed at the end of the sentences. Less clear are the results of research investigating how negation is captured in the embodied representations created during language comprehension. One possibility is that negation triggers very specific embodied effects, leading for instance to a below baseline activation of the networks responsible for representing the respective experiential dimensions mentioned in the sentences. Another possibility is that negation is captured in the sequence of activations that the sentences give rise to. Clearly, more research is needed before definite conclusions regarding the embodied nature of negation representation can be drawn. Future research should also continue to investigate the question to what degree negation processing involves general cognitive mechanisms that are also utilized in non-linguistic cognition, such as in memory or action control.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ......................................................................................................................

   ......................................................................................................................

    

.. I

.................................................................................................................................. D of research have illuminated the syntactic and semantic licensing requirements on negative polarity items (NPIs), but relatively little is known about how these licensing requirements are satisfied in real time as sentences are interpreted. Here we pursue the strategy of using the error profile of the online NPI licensing mechanism to provide clues about the inner workings of that mechanism. This is the logic underlying the study of grammatical illusions, cases where native speakers experience a fleeting perception of acceptability or unacceptability that mismatches their more considered judgments. The selective appearance of these illusions has proven useful in several domains of sentence processing research such as the study of agreement attraction, for instance *The key to the cabinets are on the table. In this chapter, we review the findings on negative polarity illusions, their parallels (and, in some cases, non-parallels) with other grammatical illusions, and the implications of this line of research for understanding the incremental processing of negative sentences as well as negative polarity phenomena more broadly.

... NPI illusions: Basic profile Negative polarity items such as any, ever, yet, or lift a finger are licensed when they occur in the scope of negation or similar operators such as no, not, few, rarely, or doubt, often described as the class of downward-entailing operators (Ladusaw ). Hence the NPI ever is licensed in () because it is in the scope of the main clause subject no bill. It is not licensed in (), because the sentence includes no licensor. Nor is it licensed in ()—although the potential licensor no senators is present in the sentence, it fails to take scope over the NPI because it is embedded inside a relative clause modifier of the subject. NPI illusions, first demonstrated in German in Drenhaus, Saddy, and Frisch (), involve the fleeting perception of acceptability of an unlicensed NPI in a sentence that contains a potential, but structurally inappropriate licensor, such as ().

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   ()

No bills that the senators voted for will ever become law.

()

*The bills that no senators voted for will ever become law.



() *The bills that the senators voted for will ever become law. When speakers have no time limitations on providing judgments, sentences like () and () receive comparably low ratings, whether the task involves binary or gradient judgments. In other words, linguists’ claim that the potential licensor in () is irrelevant to licensing of ever is readily confirmed by large-scale judgment studies. The contrast in acceptability between () and () emerges in measures involving faster responses, including the speeded acceptability task, in which comprehenders read sentences at a fixed presentation rate and then have just a couple of seconds to respond with a binary acceptability judgment. In this task sentences like () are typically accepted in a significantly larger proportion of trials than (), though not as frequently as () is accepted. Acceptance rates vary across studies for a variety of reasons, but typical acceptance rates in a speeded acceptability task are –% for (), –% for (), and –% for (). This suggests that, at an early stage of interpretation, the ungrammaticality of () is less apparent than the ungrammaticality of (). This discrepancy is taken to indicate the susceptibility of the NPI-licensing computation to errors under particular circumstances. Much work has now gone into trying to figure out what those circumstances are, with the expectation that this will illuminate the normal mechanisms of online NPI licensing. Since the NPI illusion was discovered, it has been shown to be robust across methods and languages. The effect has been demonstrated in German (Drenhaus, Saddy, and Frisch ), English (Xiang, Dillon, and Phillips , ; Parker and Phillips ; Ng and Husband ; de Dios Flores, Muller, and Phillips ), Turkish (Yanilmaz and Drury ), and Korean (Yun, Lee, and Drury ). In addition to speeded acceptability, the illusion has been observed in self-paced reading, eye-tracking, and event-related potential (ERP) experiments. In each of these cases the illusion appears as a reduction in the disruption otherwise associated with encountering an unlicensed NPI as in (). Two points should be highlighted. First, NPI illusions are not defined prescriptively, as a divergence between individual judgments and population norms. They are diagnosed based on a divergence between speakers’ considered judgments and those same speakers’ speeded responses. Second, although the illusions have been found across multiple languages and experimental measures, the cross-linguistic diversity of NPIs and licensing environments is such that we should be cautious about assuming that similar illusions will be found for all languages, all NPIs, or all configurations.

.. S  

.................................................................................................................................. A few initially appealing hypotheses about the source of NPI illusions warrant discussion. Throughout most of this chapter we assume that NPI illusions offer a window into the computations that underlie normal, successful licensing of NPIs. However, there are of course many component processes to sentence processing and any one of these could, in

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



    

principle, give rise to interpretive errors. Below we sketch a few proposals along those lines and why they are unlikely to be sufficient. One hypothesis that we often encounter when discussing NPI illusions is that speakers erroneously judge sentences like () as acceptable because of an error in signal detection. Although ever in its position in () is clearly unacceptable when carefully considered, and clearly ungrammatical under most theories, the orthographically and phonologically similar never is grammatical in that position. On some proportion of trials, then, participants might simply mis-hear or mis-read the input. While there is some independent evidence that related mis-perceptions occur in other cases (Levy et al. ), this explanation suffers two key drawbacks. First, it is unclear why () but not () suffers this problem, as substituting never would make the sentence acceptable in either case. Second, de Dios Flores () showed that although the substitution of never in () is indeed grammatical, speakers often find it difficult to process because of the close proximity of two negative words (no and never), and thus reject these sentences on a large proportion of trials. If NPI illusions reflect substitution of never for ever, then speakers should encounter at least as much difficulty with () as with (), contrary to fact. Another idea that we commonly encounter is that NPI illusions reflect the presence of multiple locally coherent substrings. For example, (), like (), gives rise to illusions. The suggestion is that (), unlike its baseline (), contains several locally coherent strings: the authors that no critics recommended for the award is, on its own, a perfectly well-formed subject, and no critics recommended for the award have ever received acknowledgment for a best-selling novel is, on its own, a perfectly well-formed sentence with a reduced relative clause (it is now the critics who are being recommended for the award). Perhaps the fact that this substring has a parse with a well-formed NPI is the source of NPI illusions. ()

*The authors that no critics recommended for the award have ever received acknowledgment for a best-selling novel.

()

*The authors that the critics recommended for the award have ever received acknowledgment for a best-selling novel.

Again, the parallels with other findings in the sentence processing literature make this hypothesis initially appealing (e.g. Tabor, Galantucci, and Richardson ). However, the compatibility of () with a reduced relative clause parse is a quirk of this and a few other examples, not a general property of the sentences that yield NPI illusions. For example, in () the equivalent substring would be no senators voted for will ever become law. This may technically be grammatical, but it reads with the difficulty of a garden path sentence such as The horse raced past the barn fell. Comprehenders are not generally known to resort to reduced relative clause parses as a strategy for getting out of difficulty. The sentences used to demonstrate NPI illusions in German are even less amenable to this explanation. The substring in () is wholly ungrammatical. ()

*Ein Mann, der keinen Bart hatte, war jemals glücklich. a man who no beard had was ever happy ‘A man who had no beard was ever happy.’

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   ()



*Keinen Bart hatte, war jemals glücklich. no beard had was ever happy

Since the local coherence hypothesis predicts NPI illusions in only a subset of the sentence types where they have been observed, we do not consider it a viable explanation. A third hypothesis that we often encounter is that the error in detecting the ungrammaticality of sentences like () is related to the scope of the negative quantifier. If the quantified subject of the relative clause takes scope in the main clause via quantifier raising, then a c-command relation between licensor and NPI may obtain (at some level of representation). This hypothesis comes in many flavors, including suggestions that (a) quantifier scope is inherently hard to compute, so participants might sometimes just guess when judging the acceptability of sentences like (); (b) sentence () is in fact grammatical because the quantifier takes wide scope, but it is difficult to process; (c) although a widescope interpretation of the quantifier in () is ungrammatical, quantifiers do sometimes take exceptional scope, and so the parser entertains this hypothesis for a brief period of time. Of these, we consider (c) the only potentially viable version, and we return to it in section ... Options (a) and (b) are unlikely explanations for the following reasons. If the presence of a quantifier adds processing difficulty, leading speakers to guess randomly when judging the acceptability of sentences like (), we should expect that all sentences containing quantifiers should receive an acceptance rate closer to %, relative to non-quantificational controls, including fully well-formed sentences. However, acceptance rates for the grammatical control condition in () are typically near ceiling. The mere presence of a quantifier does not cause speakers to guess more frequently than they otherwise would. One might accommodate this objection by narrowing the hypothesis to quantifiers in embedded clauses. However, the filler items in NPI illusion experiments often include embedded negative quantifiers, to prevent participants from adopting unnatural parsing strategies, and we do not observe reduced accuracy for these control sentences. Furthermore, this hypothesis only straightforwardly predicts illusions in speeded acceptability studies; it is not clear why or how a guessing strategy that is prompted by the presence of a quantifier should lead to an effect on reading times or ERPs at the NPI. If NPI-illusion type sentences are, in fact, grammatical due to the possibility of scoping the quantifier out of the relative clause, but result in degraded acceptability due to parsing difficulty, we might expect slower, more careful judgments to show an even greater likelihood of acceptance. This is not the case. Typically the longer one thinks about (), the worse it sounds. Of course, there are other such cases in the literature—multiple center embeddings are a classic case of a grammatical but unacceptable structure, and such sentences never seem to become acceptable, regardless of time constraints. One can, however, figure out what center embeddings would mean if they were acceptable. The same cannot be said of NPI illusions. We know of no formal investigation into what interpretation participants arrive at when they accept these sentences, but a quantifierraising explanation makes some predictions: the meaning of () should be, approximately, ‘it is not the case that there exists some x such that x is a senator and the bill that x voted for will ever become law.’ This does not seem to align with what () means, if indeed () means anything at all. A final concern is that if raising a negative quantifier out of a relative clause is grammatical, we should expect to see other reflexes of these interpretations, such as the

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



    

ability to bind a pronoun in the main clause, contrary to fact (see Kush, Lidz, and Phillips , Experiment a). For these reasons we think it unlikely that these sentences are in fact grammatical due to quantifier scope. We thus find no strong evidence to support claims that NPI illusions arise due to problems in representing the perceptual input, problems in ruling out a locally coherent parse, or (some) problems of quantifier scope. Explanations that place blame on the online NPI licensing mechanism, which will be discussed in section ., seem more plausible.

.. T  

.................................................................................................................................. One additional clarification is necessary before these hypotheses can be considered. The following explanations take for granted that NPI illusions reflect failure to detect a grammatical error, but it is in principle possible that the process is one of successful error detection followed by repair. Under the first scenario the licensing mechanism itself is error-prone and may allow ungrammatical structures, while under the second scenario it is not, and instead the repair strategy allows ungrammatical structures. Either scenario could lead to the observed elevated end-of-sentence acceptance rates. Xiang, Dillon, and Phillips () investigated this issue using ERPs, which are wellsuited to this question because of the extremely fine-grained time-course information they provide. If the illusion profile observed in behavioral data reflects a repair process that is initiated only after the violation is successfully detected, we should expect to see a stage in the ERP response when the intrusion condition and ungrammatical baseline pattern together, to the exclusion of the grammatical baseline. This is not the case. Rather, the earliest point in time where the ungrammaticality of () is recognized (i.e. the earliest point where the ERP response to this condition diverges from the ERP response to sentences like ()) is the same as the earliest point where the illusion arises (i.e. when () and () diverge). This finding suggests that for NPI illusion sentences like (), there is no early stage at which the ungrammaticality is detected. We now turn to the two classes of hypotheses that have received the most attention in the literature on NPI illusions: accounts that place the blame on the properties of the memory architecture engaged in resolving all long-distance dependencies, and accounts that place the blame elsewhere, typically in interpretive mechanisms.

... Memory mechanisms Vasishth et al. () assume that NPI licensing involves retrieving a licensor in memory upon encountering an NPI, and they propose that NPI illusions reflect noisy memory retrieval processes, motivating this claim using modeling evidence in the ACT-R framework. We now take a quick detour to describe the memory architecture that is assumed. Elements in memory, including terminal nodes of a tree (roughly, words) and nonterminals, are stored as ‘chunks’ or bundles of feature-value pairings. The relationships between nodes are encoded as features on those chunks, so that the representation of a sentence is a collection of chunks, each of which encodes its links to other chunks. Those

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



chunks can then be retrieved from memory based on retrieval cues that match the features on the chunks. Retrieval success is a function of both feature match and the chunk’s level of activation. A memory access process involving more than one retrieval cue can lead to a partial match when a chunk in memory matches some but not all of the feature specifications of the retrieval cues. Multiple partial matches can potentially occur for any given attempt at retrieval. Partial matches play a key role in this account of NPI illusions because a partial match can lead to successful retrieval as long as the activation of the chunk is sufficiently high. ACT-R is a computational model of human cognition that implements a version of this theory of memory access (Lewis and Vasishth ). Applying this to NPI illusions, Vasishth et al. characterize the computation that leads to the misperception of acceptability as retrieval from memory of the structurally irrelevant licensor, which occurs because of that chunk’s partial match with the retrieval cues on the NPI. They use +c-commanding and +negative as the retrieval cues. C-command is a structural notion that refers to the relation between a node and its sister, and any node contained within the sister. It is closely related to the notion of logical scope. Vasishth et al. acknowledge that a more developed theory should involve a better understanding of the structural cue than simply labelling it as c-command, but they leave that issue for future development. With these retrieval cues, on some proportion of trials retrieval yields success via a partial match with the non-c-commanding negative DP, and because the search was successful, no error signal indicates to the comprehender that the sentence is ungrammatical. Note that this hypothesis relies on very general principles of memory and retrieval processes, and so it predicts that illusions should occur whenever a dependency resolution mechanism gives rise to partial matches. In fact, similar explanations have been proposed for agreement attraction, an illusion of subject-verb agreement (Wagers, Lau, and Phillips ). This predicted generality is not, in fact, borne out; we will see in sections . and . that NPI illusions are surprisingly constrained in their profile. Amending the memory-based account to accommodate the specificity of the illusion is not straightforward.

... Interpretive mechanisms The other prominent hypothesis, or family of hypotheses, that has been proposed is that NPI illusions arise because of errors in the processes by which NPI-containing sentences are interpreted. The details of what computations are involved in interpreting NPIs and how those computations go wrong vary between accounts.

.... Quantifier scope As we discussed above, one such hypothesis places blame for the illusion in the interpretation of quantifiers. In an NPI illusion sentence, the licensor is typically a negative quantifier and that quantifier is unable to properly license the NPI because of its structural position inside of the relative clause. However, it is well known that the interpretation of a quantifier does not always match its surface syntactic position. This is demonstrated by the ambiguity of sentences like (), and the possibility of (), in which the pronoun it seems to be bound by the quantifier every although every is inside a nominal modifier.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

 ()

     Every girl likes some boy. a. For every girl x there is some boy y such that x likes y (each girl may like a different boy) b. For some boy x, every girl y is such that y likes x (all the girls like the same boy)

()

[Someone from everyi city] hates iti.

(May )

Since a quantifier’s scope is not determined by its surface position, the sentence comprehender may, upon encountering a quantifier, initially consider all in-principle scopal possibilities. The brief consideration of a wide-scope interpretation of the quantifier in an NPI illusion sentence leads to the initial perception that the NPI is in fact within the scope of the quantifier. As we will see in section .., this hypothesis makes some accurate predictions regarding intrusive non-quantificational licensors. However, pinning NPI illusions on quantifier scope may be problematic since (a) negative quantifiers are not able to take exceptionally wide scope in the ways shown by () and (), and (b) wide scope out of a relative clause is not generally possible, and NPI illusions have typically been shown using relative clauses. If the parser has access to these two grammatical facts, it should never consider a wide-scope interpretation of the quantifier in NPI illusion sentences. There would have to be a strong enough bias for wide-scope interpretations, in general, that the online comprehender disregards these grammatical facts. We know of no evidence for such a bias.

.... “Rescuing” by contrastive implicatures One concrete proposal is due to Xiang, Dillon, and Phillips (), who argue that (a) successful NPI licensing can be driven by negative pragmatic inferences, and (b) in the sentences that typically yield NPI illusions similar negative inferences can sometimes arise. Under some theories of NPI licensing, emotive factives like surprised license NPIs because they license negative inferences, although they do not explicitly encode negativity (Giannakidou b). For example, () licenses the inference in (). () clearly places the NPI within the scope of negation. The claim is that the close relationship between () and () allows the NPI in () to be “rescued.” ()

I’m surprised we have any sugar.

()

I thought we didn’t have any sugar.

The motivation for a separate pragmatic rescuing operation for NPIs comes from both the theoretical literature on NPI licensing, particularly comparison of English and Greek (see Giannakidou b for an overview) and ERP findings suggesting a qualitative difference between NPIs licensed by emotive factives and those licensed by more overtly negative licensors (Xiang, Grove, and Giannakidou ; see Giannakidou and Etxeberria  for further discussion of these findings). NPIs licensed by no, only, or very few yield a difference in both N and P amplitude when compared to NPIs with no licensor, whereas NPIs licensed by emotive-factives only differ from unlicensed NPIs in the N time window.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  



This could reflect an additional reanalysis process, possibly pragmatic in nature, which occurs for NPIs not preceded by a canonically negative licensor. Xiang et al. argue that a similar comparison to pragmatic inferences is responsible for NPI illusions. Sentence () is thought to give rise to an implicature. Since the speaker has specified a particular set of bills—those that no senators voted for—it can be assumed that the speaker does not have warrant to make the more general claim in (). Thus there is a suggestion that there are some other bills—those that the senators did vote for—which would make () false, and so we get the inference from () to (). Since () places the NPI within the scope of negation, the same mechanism that is responsible for NPI rescuing in () can apply. ()

The bills that no senators voted for will (ever) become law.

()

The bills will (ever) become law.

()

The bills that the senators voted for will not (ever) become law.

The specifics of when a negative inference leads to true grammaticality and when it leads to an illusion are not spelled out. A difficulty for this hypothesis lies in defining the circumstances that give rise to these inferences such that they are not predicted to occur for ungrammatical baseline conditions, which also contain relative clauses that could give rise to contrastive negative inferences. A further complication is that Xiang, Dillon, and Phillips () found that unlicensed NPIs and NPIs in illusion configurations differ in the P time window, which is unexpected under the interpretation that the P reflects pragmatic reanalysis of unlicensed NPIs. We return to some of these issues in section ...

.... Covert exhaustification The hypothesis that pragmatic inferences lie at the heart of the interpretive mistake is only one of a family of hypotheses that highlight the role of interpretation in NPI illusions. Another version of this idea is the hypothesis that NPI illusions arise because comprehenders infer a covert exhaustification operator which dominates the NPI (Mendia, Poole, and Dillon ). The overt exhaustification operator only is in fact an NPI licensor, so the thinking is that the silent version might be considered as a possible licensor. Some evidence in favor of this hypothesis comes from the relative increase in acceptability of ever in sentences with no clear licensor, but a contextually driven bias toward exhaustive interpretations as in (), compared to those without a bias toward exhaustive interpretations as in (). ()

Whenever the summer is really dry, Susy expects all of her plants to die. However, a small number of the plants have ever died.

()

Whenever the summer is really rainy, Susy expects none of her plants to die. However, a small number of the plants have ever died.

As with the contrastive inference hypothesis, a difficulty for this theory lies in identifying the factors that drive the comprehender to infer the exhaustive operator, such that they occur for illusion sentences but not for ungrammatical baseline sentences.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



    

.... Licensing as integration into context One final proposal is that illusions arise because of problems in rapidly transitioning out of the relative clause, which has a downward-entailing and NPI-compatible meaning. Under this hypothesis, online NPI licensing is not implemented by means of retrieval of an individual licensor but rather by means of integration of the semantic content of the NPI into a phrasal meaning, which results in successful integration only if the phrasal meaning has the appropriate characteristics, such as downward-entailment. It has been suggested that the contribution of an NPI is to strengthen a negative claim (Kadmon and Landman ). When an NPI is encountered soon after strong negative claims are suggested in the relative clause, the comprehender may mistakenly integrate the NPI into this contextual meaning, despite syntactic information that signals that this is unallowed. This hypothesis correctly predicts the two most startling restrictions on NPI illusions— their sensitivity to intervening material and their disappearance with licensors that merely allow but do not encourage strong, exceptionless inferences at the clause level. We now turn to these surprising patterns and their implications for the hypotheses discussed so far.

.. R  NPI 

..................................................................................................................................

... Licensor distance Although the NPI illusion effect appears to be robust across measures and languages, instances where the illusion fails to arise in specific contexts are particularly informative. Increasing numbers of studies have found that the conditions under which NPI illusions arise are more specific than simply the presence of a non-c-commanding licensor. Some of the accounts discussed above predict a more even profile for the illusion than has been observed, while others predict that the phenomenon should be sensitive to factors that do not appear to strongly impact the illusion. One striking limit on the illusion is that it disappears when the lure is suitably far away from the NPI (Parker and Phillips ). Sentences with either an embedding clause or a parenthetical intervening between the intrusive licensor and the NPI were found to not yield illusions (a and a), whereas closely matched controls with less material between the intrusive licensor and the NPI (b and b) elicited clear illusions in both speeded acceptability and self-paced reading measures. Parker and Phillips argue that it is simply the passage of time that makes the illusion go away, based on the assumption that the parenthetical phrase in (a) has no impact on the relative structural positions of the lure and the NPI. (a)

*The journalists that no editors recommended for the assignment thought that the readers would ever understand the complicated situation.



P3


P6




 .. Percent correct on each condition per patient—the non-linguistic conditions.

including image analysis and statistical analysis that parse the brain’s grey matter into areas. While some of the maps, in particular of primary sensory and motor areas, seem to be similar to what are historically known as Brodmann Areas (e.g. BA, BA, and, to a large extent, BA and  of Broca’s region; Amunts et al. ), the vast majority of areas do not have similar counterparts in this historical map (Brodmann ), but provide a more detailed subdivision of the cerebral cortex. In contrast to Brodmann’s map, the JuBrain atlas considers intersubject variability as a feature to describe an area (“probabilistic”), and provides true stereotaxic information, a prerequisite for comparison with findings from functional neuroimaging. The JuBrain atlas has a spatial resolution of mm3 voxels. It is freely available to the research community and linked to other data modalities in the HBP Human Brain Atlas (). This atlas allows computational comparisons between distinct brain areas, and is a tool to evaluate voxels, or clusters of voxels, acquired by other methods in healthy subjects and patients, for example fMRI activation clusters, voxels containing focal lesions, etc. Such methods produce results whose quality is quite different from visual inspection of topographic landmarks observed in an image. When the coordinates in common reference space of a cluster’s voxels are known, they can be located by co-registration to the JuBrain atlas (cf. Amunts et al. ; Santi and Grodzinsky  for applications to fMRI language studies). Once a lesion is mapped (or masked through a difficult semi-automatic process), the anatomical addresses of all its voxels are known, and can be mapped onto the atlas (Hömke et al. ), resulting in information about the cytoarchitectonic correlates of the lesion. As lesions are caused by pathological processes that do not respect histological boundaries, we typically obtain a list of cytoarchitectonic areas that are lesioned, where each is listed with the degree to which it is compromised. Thus in Table ., % of the posterior part of Broca’s area of the left cerebral hemisphere, namely area L in the JuBrain, is lesioned for patient P; while % of the posterior part of his left Broca’s area, L, and only % of his left anterior insula Id_L, are lesioned. These sophisticated mapping tools therefore provide a quantitative picture of the patient’s brain, which can be compared to his/her impaired functions for localizing purposes.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



   .

Table .. Percent lesion in four cytoarchitectonic language areas L L Id_L TE_L

(Broca’s region) (Broca’s region) (anterior Insula) (Wernicke’s area)

P1

P2

P3

P4

P5

P6

 

   

   

   

   

   



Clearly this atlas becomes amenable to computational investigations when data from larger numbers of patients are available. The extreme difficulty in patient recruitment and scanning has left us with only six patients, all affected in the left hemisphere. While every one of these patients is also affected in other brain regions, we restricted ourselves to four language regions. In Table ., we show the extent of the lesion (% destroyed) each of them suffered in four left hemispheric brain areas, known to support language. In the absence of a larger sample, we can only use this information informally to find some generalities as well as individual differences:

... An informal analysis We have thus far reported four different classes of quantitative measures about our patients: clinical, anatomical, syntactic, and semantic, negation-related. Ideally, we would apply computational methods in order to uncover systematic relations between these measures. However, the paucity of cases provides little opportunity for quantitative analyses, limiting us to an informal discussion. Consider first the uninformative performance patterns of P and P: Table . seems to explain them. The relevant anatomical areas in the brain of P are virtually unaffected, hence his performance is indeed expected to be near-normal; the brain of P, by contrast, is affected in a sweeping fashion—both anterior and posterior language regions, including Wernicke’s area, are compromised. As Table . shows, one cannot be sure that P even understood the instructions. Her across-the-board chance-level performance is thus not unexpected. Next, consider the performance patterns of P and P. P’s syntactic comprehension is typical of Broca’s aphasia, whereas P’s is better (though not far from typical). In the polarity experiment, they were both well above chance on the symbolic conditions and on the UE linguistic conditions. On the DE conditions, P was at chance whereas P was below chance. For both, Broca’s region is seriously injured. Yet, whereas P lost his left anterior insula, % of this brain area is spared for P. It is possible (though by no means certain) that this anatomical difference accounts for their performance difference. Still, any assertion that the difference in these patients’ lesions translates directly into the measured performance difference would require a broader empirical basis, namely many more patient scores. Less clear is the relation between the lesion and the behavioral patterns of the remaining two patients: for P, who presents a typical picture of Broca’s aphasia, the anterior language region is

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



destroyed, while the posterior one—Wernicke’s area—is spared. By contrast, for P, all four language areas are almost completely wiped out. Still, he seems to present with a structured pattern, performing above-chance on both symbolic conditions, (barely) above-chance on the conditions that contain a UE quantifier, and at chance level on the DE conditions.

.. I

.................................................................................................................................. We reviewed experimental results regarding negation-related behavior in the healthy and impaired brain. We began by arguing that tightly controlled experiments with overt negation are hard to come by, and proposed an alternative method: implicit negation, hidden inside proportional, degree, and comparative quantifiers. We provided linguistic methods for the detection of this negation and showed that it incurs a processing cost—the Downward Entailing Cost or DEC effect, found in several RT experiments with a variety of implicit negations. This finding supports a decompositional approach to implicit negation generally, but specifically, in the realm of comparative constructions. The angle we proposed was wider, because our interests lie not only in the representation and processing of negation, but also its brain mechanisms. We briefly reviewed findings from imaging suggesting that (i) negation is localized in cortex; (ii) the neural tissue that support mechanisms for the processing of negation are distinct from, though adjacent to, Broca’s region. We then proceeded to describe the details of an experiment with Spanish speaking aphasic patients, that sought to establish similar conclusions through the use of five data sources: clinical diagnosis, syntax comprehension test, a hidden negation test, and anatomical lesion analysis. Here things became more complicated. The behavioral pattern we uncovered was not inconsistent with the expected one, but vague at times. Yet, while the performance patterns observed through RT and fMRI studies in health were very refined, we noted high inter-patient variability in our population, one that resulted in a coarse anatomico-behavioral pattern containing apparent contradictions. The range of data for each patient were broad, yet the number of cases was too small for firm conclusions. The limited evidence we obtained from aphasia, then, is suggestive, though not compelling. At present, the failure to find a stable relation between behavioral deficit and lesion anatomy is due to the small number of patients tested. In the past, it has been shown that patterns emerge with larger number of patients (Drai and Grodzinsky, a, b). We therefore hope that methods for large-scale testing of braindamaged patients will be developed, to enable a more solid lesion-based perspective on negation processing and related cognitive components.

A

.................................................................................................................................. Partially supported by Israel Science Foundation, Grants No. / (Y.G.) and No. / (Y.L.), The Gatsby Charitable Foundation (Y.L.) and by the European Union’s Horizon  Research and Innovation Programme under Grant Agreement No.  (HBP SGA) and No.  (HBP SGA). Contact: [email protected].

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



   .



.................................................................................................................................. M       An early study (Carpenter et al. ), carried out with materials developed in psycholinguistics (Clark and Chase ) took no steps to control for the number of words, and merely contrasted activation for negative sentences that depict spatial relations between objects, and their affirmative counterparts: (A)

a. It is not true that the star is above the plus. b. It is true that the star is above the plus.

Later fMRI studies probed negation while attempting to set up proper controls. In a    study of double negation with complex sentences, Bahlmann et al. () similarly controlled the appearance of a negation word with another. The German nicht was featured in the matrix clause and/or in an embedded clause. When in the former, it was controlled by schon (which the authors translated as indeed), whereas in the embedded clause it was controlled by wirklich (translated as really): (B)

a. Es ist nicht wahr, dass Peter Thomas letzte Woche für das Projekt nicht einstellte.

It is not true, that Peter did not hire Thomas last week for the project. b. Es ist schon wahr, dass Peter Thomas letzte Woche für das Projekt wirklich einstellte.

It is indeed true, that Peter really hired Thomas last week for the project. Like Tettamanti et al.’s study, Bahlmann et al. feature expressions that carry non-negligible semantic weight (schon and wirklich introduce a host of presuppositions). Here, too, semantic equivalence is not maintained.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

  ......................................................................................................................

                 Implications for bilinguals ......................................................................................................................

 . 

.. I

.................................................................................................................................. T goal of this chapter is to outline relevant issues and questions regarding how bilingual individuals perceive negation in sentences. To paraphrase Klima’s seminal work (Klima ) it is not inappropriate for a work on negation (in bilingual populations) to start with negative statements. First, acquisition of negation in bilinguals will not be considered; rather we focus on neural correlates of sentence perception by adult balanced bilinguals, that is, bilinguals proficient in both languages, who use their languages daily (Li ). By limiting the type of bilingual population under review, the number of questions at hand are necessarily limited, and potentially answerable. Second, this work will not review current models of online bilingual sentence processing, since this has already been done elsewhere (see Kotz ; Kroll and DeGroot ); furthermore, none examine negation.¹ Third, the conventional question associated with bilingual individuals and their requisite neural architecture will not be asked, where the form of such a question invariably manifests as: “Does the brain of a bilingual process negation differently from that of a monolingual?” The reason this question is not under consideration is because it is unclear what could be learned from it. That is, it is already evident, prima facie, that bilinguals are not monolinguals. Therefore, the question should not be whether their neural architecture differs but how does it differ from the neural architecture associated with monolinguals. And there is the rub. The problem is that it is now well known that monolinguals display significant variation in how they process sentences (Kidd, Donnelly, and Christiansen ). As such,

¹ For an important exception, see work by Lipski (a, ).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 . 

an understanding of the systematic variation observed in behavioral and neural responses to sentence perception in monolinguals is fundamental for studies in bilingual sentence processing. In this chapter, we present our model of sentence processing, ‘Heuristic first, algorithmic second’, which was developed by examining scope ambiguous sentences. We claim this conception of the human sentence processing mechanism can clarify (i) previous findings regarding individual differences in scope interactions between negation and quantified Noun Phrases (NPs) in English and Korean, (ii) previous findings regarding neural correlates during sentence perception of sentences exhibiting negation, (iii) and finally, ways of elucidating neural responses associated with bilingual sentence processing. In this work, experiments on scope ambiguity from our lab and elsewhere will be reviewed, and it will be argued that an understanding of previous findings on the scope of negation can best be understood once individual differences regarding quantifier scope interpretation are taken into account. Specifically, a sentence processing theory that assumes that, in real-time, people understand sentences heuristically first, rather than deeply interpreting sentences using algorithmic rules of grammar, will be argued for. That is, people are satisfied with ‘good enough’ interpretations of sentences (Caramazza and Zurif ; Christianson et al. ; Christianson, Luke, and Ferreira ; Ferreira ; Gigerenzer ; Sanford and Sturt ; Simon ; Swets et al. ; Townsend and Bever ). Consequently, the rules of quantification that occur at the level of Logical Form (May ) are not necessarily immediately used by the sentence processing mechanism. However, if required by task or context, these rules can still be accessible, and applied. As such, the model is called ‘Heuristic first, algorithmic second’.² These ideas will be explored in this chapter, especially how these are fundamental to our understanding of negative quantifier scope in bilingual contexts.

. T : N   

.................................................................................................................................. Sentences such as (a,b) vs. (c,d) show that the negative operator is sensitive to scope interactions. In (a,b), the answer to the question posed is ‘yes’ for (a) and ‘no’ for (b). In contrast, the addition of the universal quantifier in (c,d) changes the potential answer to the question in (d). Whereas it is still ‘yes’ for (c), now it can be either ‘yes’ or ‘no’ for (d): ()

a. Men enjoy washing dishes. Do men enjoy washing dishes? b. Men do not enjoy washing dishes. Do men enjoy washing dishes? c. All men enjoy washing dishes. Do all men enjoy washing dishes?

² This model differs from other ‘good enough’ approaches, as cited above, in a very specific way: it is explicit about the timing associated with heuristic vs. algorithmic processes; where the latter is secondary, and that too, dependent on task/stimuli. For further details regarding the model, the reader is referred to Dwivedi ().

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

       d. All men do not enjoy washing dishes. Do all men enjoy washing dishes?



(Fodor , adapted from Horn )

The scope of negative operators is also relevant regarding negative polarity items (NPIs) such as any and ever. NPIs typically must occur in the scope of negation (Collins, Postal, and Yevudey ; Klima ; Lahiri ; von Fintel ; see examples in English, modified from von Fintel , Ewe from Collins, Postal, and Yevudey , and Korean from Han et al. ; Hindi-Urdu from Lahiri ; see also Dwivedi ):

()

English a. I don’t think we have any salt. #I think we have any salt. b. I don’t think there will ever be another Obama. #I think there will ever be another Obama. Ewe

() a. Kofi me-kpɔ ame aɖeke o Kofi 1-see person any 2 ‘Kofi didn’t see anybody.’ b. #Kofi kpɔ ame aɖeke Kofi see person any Korean ()

John-un amwukesto an mek-ess-ta. John- anything  eat-- ‘John didn’t eat anything.’ Hindi-Urdu

() a. * koii bhii aayaa anyone came ‘Anyone came.’ b. koii bhii nahiiN aayaa anyone not came ‘No one came.’ The above examples suffice to establish that negation is a scope bearing operator that interacts with the scope of quantifiers. Furthermore, this property is exhibited crosslinguistically across languages that are typologically different and not historically related.

. S   H ,  

.................................................................................................................................. In this next section, the model of ‘Heuristic first, algorithmic second’ sentence processing mechanism is proposed, which is based on stimuli exhibiting quantifier scope ambiguity.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 . 

... Scope ambiguity Sentences of the form Every horse didn’t jump the fence exhibit scope ambiguity. Two possible interpretations for the previous sentence are either that none of the horses jumped (every > not), or out of a group of say three, just two horses jumped (not >every). The former reading corresponds to the surface scope interpretation (on a reading where, for every horse, it is not the case that it jumped the fence), vs. the latter interpretation, which corresponds to the inverse scope reading (on an analysis where, it is not the case that every horse jumped the fence. For a review of quantifier scope and negation, see Ruys and Winter , as well as Beghelli and Stowell ; Büring b; Horn ; Jackendoff , among others). Before discussing the processing of such sentences, we will first introduce a recent sentence processing model that is in fact based on empirical investigations of scope ambiguous sentences. We review our model and the relevant experiments and stimuli below. Our previous work examined the interaction between the universal and existential quantifier in sentences such as Every kid climbed a tree. The ambiguity here is that either it is the case that several trees were climbed (on a reading where, for every kid, there is a tree, such that the kid climbed it) or just one tree was climbed (on an analysis where, there is a tree, such that every kid climbed it). The former reading is the surface scope reading, and the latter is the inverse scope interpretation. Quantifier scope ambiguous sentences are interpreted according to the order in which the quantifiers every and a are interpreted. The surface scope reading (which is consistent with the surface linear order of the quantifiers in the sentence) is associated with the inference of several trees; henceforth called the plural interpretation.³ In contrast, the inverse scope reading (where the order of interpretation of the quantifiers is the inverse of linear order) is associated with the singular interpretation of tree; henceforth called the singular interpretation. The logical representation of surface scope vs. inverse scope representations are shown below: Surface scope ()

a. (∀x) (x is a kid→ (9y) (y is a tree & x climbed y)) [read as: ‘For every kid x, there is a tree y, such that x climbed y’] Inverse scope b. (9y) (y is a tree & (∀x) (x is a kid → x climbed y)) [read as: ‘There is a tree y, such that for every kid, x, x climbed y’]

On one common syntactic account, quantifier scope order preference is represented via quantifier raising, a syntactic movement operation assumed to occur at the level of Logical Form (Diesing ; May ). If language processing were assumed to operate on a ‘syntax first’ approach, algorithmic computation of quantifier scope would be expected as the driving principle for interpreting scope ambiguous sentences. ‘Syntax first’ approaches assume that sentences are immediately analyzed according to their grammatical structure, without input from other sources of information, such as word-based knowledge of realworld events, prosodic constraints, etc. (Ferreira and Clifton ; Frazier and Fodor ; ³ The plural interpretation is also known as/consistent with the distributive reading, in semantic theory.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

      



Frazier and Rayner ). On this account, the interpretation that is preferred is the one that requires the least amount of structure to build (as in the Minimal Attachment hypothesis (Frazier and Fodor ; see also the Minimal Structure Hypothesis in Dwivedi ) which would be the surface scope/plural interpretation. Although this mechanism of interpretation has often been called ‘syntactic’, the more general term algorithmic computation will be used here, especially given that we have argued elsewhere that semantic computation is not independent of grammatical considerations (Dwivedi et al. a; Dwivedi ; Dwivedi et al. ). The surface scope interpretive preference of scope ambiguous sentences was empirically supported in a study by Kurtzman and MacDonald () (see also Anderson ; Fodor ; Tunstall ). In an end-of-sentence online acceptability task, they showed that participants judged plural continuation sentences as more acceptable, vs. singular continuations, after reading quantifier scope ambiguous context sentences. The experimental paradigm and findings are summarized in Table .; ?? indicates the judgment that a continuation sentence is less acceptable following a context sentence, whereas ✓ indicates a preferred continuation. In Table ., the singular continuation is shown as less acceptable following ambiguous contexts than it is following unambiguous contexts, a pattern associated with an inverse scope interpretation. In contrast, the plural continuation is acceptable following both ambiguous vs. unambiguous control contexts, which is the pattern associated with surface scope. This pattern of findings is consistent with an algorithmic interpretation. Interestingly, these empirical effects have not been fully replicated in several other studies, especially as measured via plural continuation sentences (Filik, Paterson, and Liversedge ; Kemtes and Kemper ; Paterson, Filik, and Liversedge ). One potential reason why findings have been equivocal is that previous studies assumed that only algorithmic processing applied in scope ambiguous sentences. However, in an earlier Event Related Potential (ERP) study, Dwivedi et al. (b) argued that algorithmic processing did not apply to scope ambiguous sentences. Brain wave responses to continuation sentences, both plural and singular, patterned alike, indicating no preference for interpretation. In contrast, an overall bias for continuation sentences that were plural (consistent with surface scope interpretation) was found in a paper-and-pen norming study of the  scope ambiguous context sentences used in the ERP study (for further details, see Dwivedi et al. b). Following up on the ERP work, Dwivedi () conducted three self-paced reading studies, examining the effect of interpretive questions (not used in the previous ERP

Table .. Example quantifier scope stimuli Ambiguous context

Unambiguous control context

 

Every kid climbed a tree. ✓The trees were in the park. ( )

Every kid climbed a different tree. ✓ The trees were in the park.

 

Every kid climbed a tree. ?? The tree was in the park. ( )

Every kid climbed the same tree. ✓The tree was in the park.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 . 

experiment) on scope ambiguous sentence processing; in addition it examined how heuristic bias affected interpretation. Whereas the previously reported norming study showed an overall preference for plural continuations of scope ambiguous sentences, in Dwivedi (), a by-items analysis showed that some of these sentences (e.g. Every kid climbed a tree) showed a plural continuation preference % of the time, whereas others were at rates closer to % (e.g. Every jeweler appraised a diamond). Given that the structure of these sentences was identical, we posited that it was word-based information regarding the number of participants in a familiar scene/event that determined meaning (i.e. lots of kids tend to climb lots of trees but no such bias is prevalent for scenes involving jewelers and diamonds; see also Dwivedi, Goertz, and Selvanayagam  for further evidence of this hypothesis). Stimuli were then separated into separate experiments by bias: Experiments  and  examined heavily biased stimuli whereas Experiment  examined processing of unbiased stimuli. Furthermore, Experiments  and  also included questions, whereas Experiment  did not (and will not be discussed further). Results indicated that reading times (RTs) for continuation sentences (as in Table .) never showed the expected pattern of results on an assumption of algorithmic processing; that is, singular continuation sentences did not take significantly more time to read after ambiguous vs. unambiguous contexts. Instead, both Experiments  and  showed on-line RT patterns for continuation sentences that were consistent with the heuristic bias of the scope ambiguous context sentence. The only place where an effect of algorithmic processing was found was in question-response accuracy rates. That is, when participants were specifically queried about number interpretation, as in How many trees were climbed (responses were either ONE or SEVERAL), singular continuations showed significantly lower accuracy rates after ambiguous vs. unambiguous conditions—the expected pattern of results when algorithmic computation applies. Moreover, these inverse scope conditions produced response accuracies at less than chance. This pattern was found for both Experiments  and . Thus, we posited that algorithmic scope computation did in fact take place, but not during initial online reading, only later when required due to the question–answer task. As such, a ‘Heuristic first, algorithmic second’ processing architecture was posited, which clarified how the heuristic and algorithmic systems are coordinated (see relevant work on good enough approaches by Ferreira and colleagues, Ferreira ; Ferreira, Bailey, and Ferraro ; Karimi and Ferreira ; see also Gigerenzer ; Townsend and Bever ).

. I     .................................................................................................................................. ... Individual differences in quantifier scope processing Recent analyses reveal, however, that the effects reported above were driven by two different sets of participants within the studies. We re-examined the data by separating participants by their question–response accuracy rates. We found that participants with lower question–response accuracy rates for inverse scope conditions did indeed exhibit

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

      



sentence RT patterns that were consistent with an algorithmic interpretation, whereas participants with higher rates of accuracy showed RT patterns consistent with heuristic biases (Dwivedi, Rowland, and Curtiss ; Dwivedi ). Moreover, these individual differences in response accuracy and RTs were not correlated with working memory span (Dwivedi and Goldhawk ). In fact, these results regarding individual differences in quantifier scope ambiguity processing are consistent with the findings of Gibson et al. (), who presented five experiments examining sentences of the form Every tester didn’t win the game. This sentence has two possible meanings: Either it is the case that none of the testers won the game or it is the case that not every tester won, that is, some testers did win. The former interpretation is consistent with a surface scope interpretation of the sentence, where every takes scope over not; whereas the latter interpretation is consistent with the inverse scope interpretation, where not takes scope over every. The logical notation is shown below: ()

a. Surface scope (∀x) (x is a tester → ┐ ( x win game y )) ‘None of the testers won the game.’ b. Inverse scope ┐(∀x) ( x is a tester → ( x win game y )) ‘Not every tester won the game.’

Gibson et al. were following up on a series of experiments on negation and quantifier scope (Gualmini ; Lidz and Musolino ; Musolino and Lidz ) and were interested in the role that a biasing context might play for sentences such as Every tester didn’t win the game. Stimuli examples are given below in (): ()

Surface-scope-biased context: Ten novice computer gamers were asked to test a new piece of gaming software. None of them were expected to be able to win the game. Inverse-scope-biased context: Five expert computer gamers and five novice computer gamers were asked to test a new piece of gaming software. All of the experts were expected to be able to win. Ambiguous target: As expected, every tester didn’t win the game How many testers won the game? None Some All

Over the course of five experiments, Gibson et al. () found that their findings supported the existence of three different participant groups: one that was sensitive to contextual bias for surface vs. inverse scope interpretation, another that only chose surface scope interpretation, and a small group that chose inverse scope interpretation. Thus, they concluded that there are individual differences regarding whether and how people integrate meaning across sentences in discourse, and furthermore, there are also differences found at the single sentence level, when embedded in context. In other words, Gibson et al. (), like Dwivedi (), found that some individuals are heuristic only, and others are algorithmic only, regarding quantifier scope in context.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 . 

These findings have direct bearing on results regarding negation and quantifier scope reported in Han, Lidz, and Musolino (). In that work, the researchers were interested in evaluating whether conflicting theoretical findings regarding Verb movement (Emonds ; Pollock ) in Korean (e.g. no movement, see Han and Park , among others vs. verb movement, see Otani and Whitman , among others) could be evaluated via the syntax of negation. That is, they argued that facts regarding scope interaction between negation and quantified NPs could provide evidence regarding the theoretical proposal regarding Verb raising. In Korean, negation can occur as a preverbal clitic, or as a word followed by do-support, as shown below: ()

a. Motun salim-i yeki-e an o-ass-ta. every person- here-to  come--. ‘Every person didn’t come here.’ (pre-verbal clitic negation) b. Motun salim-i yeki-e o-ci ani ha-yess-ta. every person- here-to come-  do--. ‘Every person didn’t come here.’ (verbal negation)

Interestingly, Han et al. () found that sentences with quantified NPs and negation, when embedded in biasing context, resulted in different scope interpretation, divided by group. That is, as in Gibson et al. (), they observed individual differences. In addition, their numerical results echo those found in Dwivedi () regarding inverse scope, where participants performed at less than chance in both sets of studies. Upon further analysis, Han et al. divided participants into groups based on surface vs. inverse scope interpretation, and they found a bimodal distribution—confirming clear individual differences regarding scope interpretation. They concluded that their findings showed that these differences indicate that different grammars, regarding Verb movement, have been acquired by different participants for this SOV language. An alternative way of characterizing their results is proposed in this chapter, where a key difference here is that we assume that language processing should incorporate brain principles (Kutas ). We propose, to our mind, a simpler conclusion, which is that human brains differ in terms of their processing and computational strategies. Whereas all human brains have two systems of computation at their disposal—Kahneman () calls these System  and System , analogous to the heuristic and algorithmic streams discussed in Dwivedi () and elsewhere—it seems that certain human brains preferentially use one strategy over the other. In any case, the fact that we see precisely the same individual differences in English, regarding exactly the same construction (in Gibson et al. ) as well as in other quantifier scope ambiguous sentence types (Dwivedi et al. ; Dwivedi ) would argue against the idea that there are two different Korean grammar acquisition possibilities regarding Verb movement. We note that English does not have verb movement (Pollock ). Thus, when two data sets of the same form show the same pattern, by Occam’s razor, the conclusion should be the same, not different. And that conclusion is this: monolingual individuals differ in terms of their sentence processing strategies (see a recent ERP study on quantifier scope in Selvanayagam et al. , which examines dispositional affect as a possible factor that modulates sentence processing strategies; see also Dwivedi and Selvanayagam, ).

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

      



.. Individual differences in neural response In this section, we move onto the neural correlates of heuristic vs. algorithmic sentence processing strategies, in terms of event-related brain potentials (ERPs). ERPs reflect voltage changes in the electrical brain activity associated with cognitive processing. This methodology is particularly useful, as it allows for an examination of the processing of naturalistic language stimuli online with very high temporal resolution (on the order of milliseconds) and adequate spatial resolution (through scalp distribution). Two well-known language ERP components will be discussed, where each is associated with distinct aspects of semantic/pragmatic vs. syntactic or grammatical processing. With regard to the former, since the seminal work of Kutas and Hillyard (, ), it has been shown that a pragmatically anomalous or unexpected word, (e.g. He spread the warm bread with socks) elicits a large N effect, a distinct negative shift in the ERP waveform that begins about  ms after the onset of the critical word and peaks at about  ms poststimulus. This is in contrast to pragmatically felicitous or expected words (e.g. He spread the warm bread with butter) where the amplitude of the N is significantly smaller. St. George et al. () have demonstrated that, in addition to sentence-internal semantic information, N amplitude is sensitive to contextual constraints across discourse (see also van Petten ; van Berkum, Hagoort, and Brown ; van Berkum, Zwitserlood, Hagoort, and Brown ). In contrast, the P is a later-occurring ERP component traditionally thought to be elicited by structural or grammatical aspects of the linguistic input (Hagoort, Brown, and Groothusen ; Osterhout, Holcomb, and Swinney ). This component has been related to the processes of revision and repair in sentence processing. Work by Kaan et al. () indicates that the P may be a measure of syntactic integration, where constructions that are more difficult to integrate produce larger P amplitudes. Kaan and Swaab (a, b) argue that the P actually represents a family of components distributed across the scalp. P activity with a posterior distribution appears to index syntactic processing difficulty, whereas P activity with a frontal distribution is related to ambiguity resolution and/or an increase in discourse level complexity. That is, frontal P activity has been claimed to signal that a preferred structural analysis can no longer be maintained and must be revised. Thus, ERP methodology yields information regarding the qualitative nature of processing. It is clear that the N is a natural candidate as a neural correlate of heuristic processing, whereas the P is a neural correlate of algorithmic processing (Dwivedi et al. ; Friederici and Weissenborn ; Kuperberg ; Severens, Jansma, and Hartsuiker ; Steinhauer et al. ; Tanner et al. ; van Herten, Chwilla, and Kolk ). Below, we discuss another empirical finding, this time in the neural realm, that our sentence processing model can account for. Namely, we argued in Dwivedi, Goertz, and Selvanaygam () that sentence interpretation can occur without deeply attending to quantifiers (on a heuristic first interpretation). Once this is assumed, classic ERP results regarding sentences exhibiting negation can be understood. An example of a heuristic processing strategy discussed in Dwivedi () is one that uses the lexical-pragmatic association of words for interpretation. Note that, in the literature, this heuristic strategy has alternatively been called the semantics processing stream, lexical

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 . 

association, semantic association, and more recently, semantic attraction (for recent examples, see Ferreira ; Kim and Sikos ; Kuperberg ). The idea is that, in the absence of any grammatical/structural information, representations regarding events can be computed by simply recognizing the lexical-pragmatic association of words alone, for example    will always be understood as an apple-eating event by a kid. The well-known processing N1 V N2 strategy (Caramazza and Zurif ; Townsend and Bever ) plays a role here. This strategy consists of recognizing that, for the most part, English sentences are structured such that the first Noun (N1) is the subject of a sentence and the one following the Verb (N2) is the object. This sort of event interpretation relies on experience with the real world; it is independent of grammatical computation. ERP language work by Chwilla and Kolk () showed that when the final word in a triplet is unexpected (e.g.    vs.   ) an N component is elicited, where this waveform is a marker of lexical-pragmatic anomaly (Kutas and Hillyard , ). Thus, even without grammatical cues, simple word triplets can result in an event interpretation (also known as a script or schema, Rumelhart ; Schank and Abelson ). Thus, the language processor can posit an event interpretation by quickly scanning incoming linguistic material via simple word recognition, without consulting detailed syntactic and semantic rules of computation. This would result in a ‘good enough’ representation, for sentences such as Every kid climbed a tree using a ‘quick and dirty’ heuristic processing strategy only (see relevant citations above). In fact, this conception of sentence processing, with its corresponding neural correlate, can account for the first published article on negated sentences employing ERP methodology (for more recent work, see Steinhauer et al. , among others). In Fischler et al. (), stimuli of the following form were examined in (): ()

(i) (ii) (iii) (iv)

True, Affirmative: True, Negative: False, Affirmative: False, Negative:

A robin / is / A robin/ is not/ A robin is / A robin/ is not/

a bird. a tree. a tree. a bird.

The neural correlates of these sentences (where participants were required to judge whether these were true or false, and measurements were taken at the final word, bird vs. tree) showed that it was not the propositional truth value that elicited effects (i.e. neural responses to false vs. true statements did not differ measurably) but rather lexical mismatch effects that did. Sentences of the form ‘A robin . . . tree’ above in (ii) and (iii) elicited N effects. These results are in fact consistent with our model of Heuristic first processing. That is, participants used heuristic processing strategies in the experiment (here, N1 V N2), which resulted in lexical-pragmatic mismatch effects for robin . . . tree. That is, as per our experimental online results above for scope ambiguous sentences such as Every kid climbed a tree, participants did not deeply attend to the negative operator in the Fischler et al. sentences either; instead they simply attended to lexical meaning, yielding the observed N effect.⁴ ⁴ Given that participants were to judge the veracity of the sentences in the experiment, it is surprising that even with such a task they would did not deeply attend to the quantifier. We note that this experiment lacked filler sentences of any kind. That is, all the experimental sentences were of the same form ‘An X is a/not a Y ’. It could be that the uniform nature of the sentences, which all exhibited a

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

      



. I    

.................................................................................................................................. At this point, we can return to the first paragraph of this chapter and point out why we would not ask the conventional question associated with bilingual sentence processing. Namely, asking the question as to how bilingual neural responses differ to sentences of a particular type (in this case, with negative operators) as compared to monolingual neural responses is not necessarily the right place to start. Or at least, it is not the right place to start unless one first has an understanding of what consists of heuristic vs. algorithmic sentence processing (especially with respect to negative operators), and what are their neural correlates. We have seen above that there are clear individual differences with respect to how quantifiers are processed (and that these differences can best be understood in terms of our model). This is relevant especially with respect to patterns of brain activity of second language processing. This is because claims have been made regarding the progression of such processing, where early learners of a second language exhibit N effects when reading ungrammatical sentences (indicating sensitivity to lexical content). Only once these learners become highly proficient do they exhibit P-like effects (Steinhauer, White, and Drury ). Steinhauer et al. argued that different neural responses are expected of bilinguals as they progress in their proficiency, where highly proficient bilinguals are sensitive to grammatical anomaly, in the same way as monolinguals, showing relevant P effects, whereas less proficient bilinguals would only show N effects to these same anomalies. In other words, the qualitatively different nature of the ERP responses is hypothesized to correlate with different levels of grammatical knowledge (where these levels are associated with proficiency) in a second language. However, Tanner et al. () showed that even in processing straightforward grammaticality violations, such as subject–verb agreement in German, a certain set of EnglishGerman individuals showed only a P effect, and another set showed only N effects. Tanner () (see also Tanner and van Hell ) followed up on this finding with a recent paper on monolinguals, arguing that in fact, some monolingual English speakers only exhibit N effects for these same subject–Verb agreement violations, whereas others exhibit P effects (see also Pakulak and Neville ). That is, monolingual, healthy, educated, young adult English participants in university settings showed systematic variability in terms of their behavioral responses to sentences, as well as their neural responses. These responses underscore the fact that individual differences cannot be ignored in language processing (Kidd et al. ). The implications for studying bilingual populations are quite clear. Individual differences in sentence processing in the control (monolingual) population need to be better understood before strong claims regarding bilinguals can be made. Furthermore, the variability observed in two qualitatively different ERP components best known for simple structure, resulted in first-pass shallow strategies. This heuristic strategy would account for the robust N effects reported in this study.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi



 . 

language processing (the N and P) must be better characterized. Presently, we claim that a ‘Heuristic first, algorithmic second’ model of sentence processing would help elucidate this characterization. Finally, we note that a core assumption in cognitive neuropsychology (and by extension, cognitive neuroscience) as laid out by Coltheart (), that the functional architecture of the brain across individuals is uniform is clearly an idealization. In order to try to begin to understand how and why the neural architecture might differ across individuals, very recent work in our lab (Selvanayagam et al. ; Dwivedi ) investigates correlating these neural differences with dispositional affect (personality) and emotional mood.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ......................................................................................................................

              fMRI, ERP, and aphasia ......................................................................................................................

 ø 

.. W  ?

.................................................................................................................................. N is a fundamental property of human language. As far as can be ascertained, all human languages have some kind of morpho-syntactic marker of sentential negation, while no animal communication system has it. As such, it is of fundamental interest not only to linguistics, but also to psychology and to cognitive science. In English, as well as in many other languages, sentential negation is marked with an adverb, such as ‘not’, and can, for example, be diagnosed in the following ways. Only a negative sentence can be followed by a negative elliptical conjunct (), and only a negative sentence allows a negative comment as a sign of agreement (): ()

He will not get it right, and neither will she!

()

He will not get it right.—No, I agree, he won’t.

These are incompatible with positive (affirmative) polarity: () *He will definitely get it right, and neither will she. ()

*He will definitely get it right.—No, I agree, he won’t.

Sentential negation has two fundamental semantic properties, truth value reversal and entailment reversal. First, if () is true, then () must be false, and vice versa. This exclusive relation may or may not hold between () and (), or between () and ():

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ø 

()

Bruce is a maniac.

()

Bruce is not a maniac.

()

Bruce is a politician.

The second property is that negation reverses the direction of entailment. A positive (affirmative) statement is upwards entailing, which means that (), the subset of fruit (the more specific concept), entails (), the superset of fruit (the more general concept), not (), which is a subset of the subset (more specific). In other words, if () is true, then () is also true, but not necessarily (). For a negative statement, such as (), on the other hand, the entailment goes in the other direction: (), the subset (specific) entails (), the subset of the subset, but not (), the superset (the general). Stated otherwise, if () is true, then () is also true, but not necessarily (). ()

Sheila wants to buy some apples.

()

Sheila doesn’t want to buy some apples.

()

Sheila wants to buy some fruit.

()

Sheila wants to buy some Golden Delicious.

Negation also affects the syntactic structure of the sentence. In generative grammar, a negative sentence has a designated structural projection, a negation phrase (NegP), situated in the middle of the clause between IP and VP (Christensen ; Haegeman ; Zanuttini ). The presence of NegP extends the overall syntactic structure and consequently increases the syntactic complexity of the clause, as illustrated in (a–b). ()

She suspects . . .

(a)

(b) CP

CP C° that

C° that

IP DP the guests I° might

I’ VP be late

IP DP the guests I° might

I’ NegP AdvP not

Neg’ Neg°

VP be late

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



.. B : P 

.................................................................................................................................. In this section, empirical evidence using behavioral methods will be used to argue that negation increases processing load. This will be shown via response times to comprehension questions as well as via error rates. Compared to a positive statement, a negative statement takes longer to process (increases response time) and is misinterpreted more often (Carpenter and Just ). In fact, response time and error rate increase linearly with the number of negative elements (Sherman ). These negative elements include negative operators, such as not, never, and no one, negative prefixes, for example un-happy and im-possible, and lexical (implicit) negation, as in doubt and ignore, which correspond to not believe and not attend to. Compare () and (). The two sentences have the same meaning, but the number of negative elements in () makes it significantly more difficult to understand. ()

No one doubted that he would not be unsuited for the director’s job.

()

Everyone believed that he would be suited for the director’s job.

Multiple negation is also one of the factors that conspire to cause misinterpretation or comprehension breakdown in linguistic illusions such as () (which means the opposite of what most people think) (Kizach, Christensen, and Weed ; Natsopoulos ; Wason and Reich ): ()

No head injury is too trivial to be ignored.

Results of priming studies suggest that negation is applied relatively late in online processing. Hasson and Glucksberg () found that up until a second after seeing the target word, both the positive and the negative version of the same sentence primed the target associated with the positive reading. That is, both () and () primed the word vicious (presumably because sharks are considered vicious). After one second, however, the positive () still primed vicious, whereas the negative () primed the word gentle. ()

My lawyer was a shark.

()

My lawyer was not a shark.

Similarly, in a sentence-to-picture matching task, Kaup, Lüdtke, and Zwaan () found that up until  milliseconds after reading a negative statement, it was mentally represented as positive, while after  milliseconds, the representation matched the negative sentence. These results suggest that a negative sentence is initially interpreted as positive, and only subsequently interpreted as negative. This is compatible with the hypothesis that a

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ø 

negative statement is derived from a positive one by adding a negative operator, in the syntax represented as a NegP.

.. N: P   ERP

.................................................................................................................................. A negative polarity item (NPI) is a word or a phrase that can only be used in a negative context (there are a few other contexts, but they are not relevant here). In () and () the NPIs are underlined: ()

He had not seen anyone (at all) (either).

()

*He had seen anyone (at all) (either).

Conversely, a positive polarity item (PPI), underlined in () and (), is an element that cannot occur in a negative context: ()

*I do not like this quite a lot already.

()

I (do) like this quite a lot already.

For an NPI to be licensed (i.e. to be grammatical and acceptable), two conditions must be met: The first one is a semantic condition stating that a licensor (negation) must be present, cf. the difference in grammaticality in ()–(). The second condition is syntactic, stating that the licensor must be accessible; the mere presence of a negative element is not sufficient. NPIs must be in the scope of negation, which means that the NPI must be ccommanded by negation. This means that an NPI must be contained inside the constituent which is the structural sister or the constituent headed by the negative operator. Compare () and (). In (), the NPI ever is licensed: The licensor no is present (condition ) and it is accessible (condition ). The subject is headed by no and ever is contained in its sister constituent. This is also why the PPI certainly is not licensed. In (), on the other hand, only the first condition is met: Negation is present, but it is not accessible as an NPI licensor because it is embedded inside the relative clause and does not take scope over the matrix clause. This intervening negation is insufficient to license the NPI ever and to block the PPI certainly in the matrix clause in (): ()

[No student [who actually likes syntax]] will ever/*certainly consult this handbook.

()

[A student [who does not like syntax]] will certainly/*ever consult this handbook.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



Electroencephalography (EEG) measures electrophysiological changes on the scalp as a function of neuronal activation. These changes can be time locked to specific stimuli or events. An event-related potential (ERP) is a brain response to specific events. Two of the best-known ERPs for language are the N and the P (Kutas and Hillyard ; Friederici ; Osterhout and Nicol ). The N is a centro-parietal negative deflection occurring approximately  milliseconds after the onset of semantically or pragmatically anomalous or incongruous words compared to control words. The size (amplitude) of the N is standardly taken to reflect the degree of semantic fit or integration cost. In (), the word sand will elicit an N compared to water, and socks will elicit an even larger effect: ()

She poured herself a nice glass of water / sand / socks.

The P, which is a centro-parietal positive-going deflection occurring approximately after  milliseconds, is often taken to reflect the syntactic integration cost of reanalysis and repair, for example in syntactic ambiguity and garden-path sentences, and syntactic violations (Friederici ; Frisch et al. ; Kaan et al. ; Osterhout, Holcomb, and Swinney ). Saddy, Drenhaus, and Frisch () investigated licensing violations of NPIs and PPIs using ERP and sentences corresponding to ()–() on  native speakers of German. In (), the NPI jemals ‘ever’ is licensed by the negative kein ‘no’, but not by the non-negative ein ‘a’. Conversely, the PPI durchaus ‘certainly’ is incompatible with the negative kein in (). () Kein/*ein man, der einen Bart hatte, war jemals froh. No/a man who a beard had was ever happy ‘No/*a man who had a beard was ever happy.’

(NPI)

() Ein/*kein man, der einen Bart hatte, war durchaus froh. A/no man who a beard had was certainly happy ‘A/no man who had a beard was certainly happy.’

(PPI)

Saddy et al. () found that both the unlicensed NPI in () and the unlicensed PPI in () elicited an N, whereas only the unlicensed PPI in () also elicited a P, see Figure .. The same effect for an unlicensed PPI (N and P) was found by Tesan, Johnson, and Crain () using Magnetoencephalography (MEG), which measured brain activation as reflected in changes in the magnetic field around the skull. Yurchenko et al. () replicated Saddy et al.’s () ERP study with twenty-six Dutch speakers using somewhat different sentences such as () and (): () Harold kon heel erg mooi tekenen Harold could really very beautifully draw maar/en zijn handschrift was niet/nook bijster leesbaar. (NPI) but/and his handwriting was not/*also at all legible ‘Harold could draw really very beautifully but/and his handwriting was not/*also at all legible.’

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ø  Map

NPI C3

NASION Fp1 Fp2 F7

F3

Fz

F4

C3

T3

T5

Cz

P3

Pz

C4

P4 O2

O1

PPI

C4

C3

PZ

P4

P3

N400

CZ

C4

PZ

P4

F8 P3

A1

N400

(Saddy et al. 2004) CZ

T4

A2

T6

–5 μV negative polar/grammatical negative polar/ungrammatical

INION

P600 0.5

s 1.0

–5 μV

positive polar/grammatical positive polar/ungrammatical

5

0.5

s 1.0

5

 .. Left panel: EEG electrode map showing the scalp viewed from above; odd numbered electrodes on the left side of the scalp, even numbered ones on the right. Right panel: The ERP components for unlicensed NPIs (middle) and PPIs (right) compared to grammatical controls (Saddy, Drenhaus, and Frisch : ). Note that negative is plotted upwards and positive downwards on the y-axis.

()

Het beleid tegen terrorisme moet worden aangescherpt The policy against terrorism must be accentuated (PPI) en/maar de partijen zijn echt/*niet nogal eensgezind. and/but the parties are really/*not rather unanimous ‘The policy against terrorism has to be accentuated and/but the parties are really/ *not rather unanimous.’

Like Saddy et al. (), Yurchenko et al. () found that the unlicensed NPI bijster ‘at all’ in () elicited an N, but unlike the previous study, the unlicensed PPI nogal ‘rather’ in () only elicited a P but no N. The two studies involved different types of negation, which may explain this; Saddy et al. used a negative quantifier corresponding to no, which may be more difficult to process than a negative adverb corresponding to not, as also suggested by Yurchenko et al. (: ). The N arguably reflects the integration cost due to the incompatibility of the polarity item and the preceding lexical context, which is the case for both NPIs and PPIs. However, as argued by Yurchenko et al. (), the mechanisms underlying the processing of the two are different. While negative polarity is clearly sensitive to the local lexical context— consider for instance () and () above—positive polarity is sensitive to the wide discourse context. In English, the adverb still is a PPI and is incompatible with local sentential negation (). However, there is no problem if negation is not local or if the sentence is globally positive, as with double negation (), or if the PPI is echoed as part of, for example, an emphatically denied verb phrase like be single in (): ()

I met Mary yesterday. She is (*not) still single.

()

I can’t believe [Mary is (not) still single].

()

Mary is still single, isn’t she? No, Mary is NOT still single.

So, the N could be taken to reflect semantic incongruity for unlicensed NPIs, whereas for PPIs, the P could reflect syntactic reanalysis. Saddy et al. (: ) argue that the

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



P for the unlicensed PPIs reflects the parser’s attempt to reanalyze the sentence and repair the structure in order to get the PPI out of the scope of negation. For the unlicensed NPIs, there is something missing, namely negation, and there is no attempt to repair, hence no P. Yurchenko et al. (: ), however, argue that as the local syntactic environment is not central to the licensing, the P could be taken to reflect discourse processing problems as the parser searches a wider context in the attempt to construct a globally positive representation. Drenhaus, Frisch, and Saddy () used ERP to investigate the effect of intervening negation in German using sentences such as ()–() (see ()–() above): () Kein man, der einen Bart hatte, war jemals glücklich. No man who a beard had was ever happy ‘No man who had a beard was ever happy.’ () *Ein man, der einen Bart hatte, war jemals glücklich. No man who a beard had was ever happy ‘*A man who had a beard was ever happy.’

(No negation)

() *Ein man, der keinen Bart hatte, war jemals glücklich. No man who no beard had was ever happy ‘*A man who had no beard was ever happy.’ (Intervening negation) The results showed that an unlicensed NPI following an intervening negation () elicited an N, but the N is weaker (the amplitude is smaller) than when there is no negation at all (); in both conditions, the unlicensed NPI also elicited a P, contra both Saddy et al. () and Yurchenko et al. (). The intervening negation is insufficient to license the NPI because it does not c-command the NPI, but it can erroneously make the semantic or pragmatic integration easier despite the lack of syntactic accessibility, hence the weaker N. The P could reflect structural reanalysis as the parser searches the structure for a suitable licensor, or indeed because these sentences are more complex due the relative clause (who had a/no beard in () and ()), and, hence, interpretation is more difficult and the integration cost is higher.

.. N: N  MRI

..................................................................................................................................

Though it is unique and universal to human language, there are still relatively few published neuroimaging studies on negation, with Bahlmann et al. (), Carpenter et al. (), Christensen (), and Tettamanti et al. () as notable exceptions.

... It is not true . . . Carpenter et al. () investigated the processing of positive and negative sentences using functional magnetic resonance imaging (fMRI), which measures changes in blood flow in the brain as a function of cortical activity (Huettel, Song, and McCarthy ). The idea

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ø 

was that negative sentences are more difficult to process than positive sentences because negation involves polarity reversal (from positive to negative). Consequently, Carpenter et al. hypothesized that negation would increase brain activation in areas associated with language comprehension, namely left posterior temporal regions and bilateral parietal regions. In a sentence-to-picture-matching task, participants were presented with sentences such as () with or without negation, and then shown a picture that either matched or did not match the sentence (a star above or below a cross). The task consisted of indicating whether the sentences were true or false. ()

It is (not) true that the star is above the cross.

The behavioral data confirmed that negation takes longer to process than positive polarity. Compared to positive sentences, the negative sentences increased activation in two regions of the brain, namely in the left posterior temporal lobe (BA , ) and bilaterally in parts of the parietal lobe (BA , , , ). However, Carpenter et al. () only scanned and analyzed these two regions of interest, which leaves the potential involvement of other parts of the brain undetected. Another factor that potentially affected the results is the difference in length between the positive and affirmative sentences due to the presence or absence of the word ‘not’. All things being equal, adding more words to a sentence makes it more complex in structure by increasing the size of one or more constituents, which has been shown to correlate with increased activation in Broca’s area (left BA , ) as well as in the left posterior temporal cortex (Pallier, Devauchelle, and Dehaene ; Zaccarella et al. ). Furthermore, the study included only eight participants, which weakens the generalizability of the results.

... Reduced access to action representations Tettamanti et al. () scanned eighteen native speakers of Italian using fMRI during passive listening to a set of sentences that varied on two factors: polarity (positive vs. negative) and concreteness (action-related () vs. abstract ()). () Adesso io/non premo il bottone. Now I/not push.. the button ‘Now, I (don’t) push the button.’ la fedeltà. () Ora io/non apprezzo Now I/not appreciate.. the loyalty ‘Now I (don’t) appreciate the loyalty.’ Tettamanti et al. () argued that negation reduces access to mental representations of negated concepts including motor representations for action verbs. Evidence suggests that language is ‘embodied’ such that action words, such as to kick, to hit, and to bite, activate the same brain circuits responsible for the execution and observation of the action described (Pulvermüller ). This includes the relevant parts of the cortical motor system involved in movement of the legs, hands, and face for to kick, to hit, and to bite, respectively. Under negation, access to these representations is reduced.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



The results showed more activation for positives than negatives in three areas, one located in the right middle frontal gyrus, one in the right middle occipital gyrus, and one subcortically in the left pallidum. However, there was no increase in activation triggered by negation. The results showed significant interactions between polarity and concreteness such that compared to positive polarity, negation reduced the activation level in posterior cingulate cortex for abstract sentences and in fronto-parieto-temporal regions for action sentences. Furthermore, negation decreased connection strengths in the left-hemisphere fronto-parieto-temporal network for action sentences. These results are in line with the hypothesis that negation reduces access to mental representations. In summary, contrary to the behavioral evidence discussed above, the results of Tettamanti et al.’s () experiment do not support the hypothesis that negation increases processing load. The activation levels are higher for positive sentences than negative sentences, and negation as such is not associated with increased activation. However, it could be argued that since the experiment did not involve any measure of the participants’ behavior, it is at least possible that some of them did not perform the passive listening ‘task’ as predicted. There is no way of knowing whether or not they actually paid attention and parsed the sentences in full, instead of arriving at a partial interpretation via ‘shallow processing’, with or without negation (Sanford and Sturt ; Sanford and Graesser ). Hence, it is impossible to tell whether the differences in brain activation are attributable to the differences in polarity. Without behavioral measures, it can also not be tested whether differences in brain activation are correlated with response time (RT); Christensen and Wallentin () reported a positive correlation between RT and activation in Broca’s area in a linguistic task, whereas Mohamed et al. () reported a negative correlation between RT and activation in bilateral occipital (visual) cortex and in left inferior parietal ‘sensorimotor areas’ in a visuomotor response-time task (clicking a button when seeing a visual cue).

... A syntactic effect Christensen () used fMRI to investigate the syntactic and semantic contrasts between negative and positive sentences in Danish. The stimulus consisted of fourteen positive and fourteen negative target sentences such as () and (). The sentences had equal length, which avoided the potential confound of sentence length discussed in section .. above: () De sejeste mænd kender og bruger også sæbe. The toughest men know and use also soap. ‘The toughest men know and use soap too.’ () De sejeste mænd kender men bruger ikke sæbe. know but use not soap. The toughest men ‘The toughest men know but don’t use soap.’ The participants (n¼) were presented with a target sentence (shown on the screen for five seconds) followed by a positive or negative probe question (shown for three seconds) that either matched or did not match the target, and the task was to answer ‘yes’ or ‘no’ by

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ø 

pressing a button. For example, () could be followed by ‘The toughest men don’t use soap’, which would be false, and the correct answer would be ‘no’. Crucially, the fMRI data was acquired during the reading of the target sentences, not during the reading and answering of the probes. Therefore, the resulting activation contrasts are likely to be due to differences in cognitive processing of negative vs. positive polarity rather than motor activity (planning and executing movement of the right index or middle finger). Importantly, the probe task ensured that participants paid attention and parsed the target sentences in full. The results showed that negation increased activation (relative to positive polarity) in the left premotor cortex (BA ) (see the left panel in Figure .). This effect is compatible with the idea that negation increases processing cost due to the added syntactic structure (NegP). It is interesting to note that the effect was not localized in Broca’s area, which has been documented numerous times to be sensitive to structural complexity (Christensen and Wallentin ; Friederici ; Makuuchi et al. ; Pallier, Devauchelle, and Dehaene ). However, a number of studies have also found increased activation in the left premotor cortex (BA //) due to syntactic complexity (Ben-Shachar, Palti, and Grodzinsky ; Bornkessel et al. ; Christensen and Wallentin ; Christensen, Kizach, and Nyvad ; Zaccarella et al. ), working memory for structured chunks or sequences (Bor et al. ; Bor and Owen ), as well as in other rule-based, non-motor mental-operation tasks (Hanakawa et al. ). This result (negation increasing activation relative to affirmation) is contrary to the result reported by Tettamanti et al. (). However, as mentioned, this may be due to the use of passive listening rather than some kind of experimental task ensuring that participants parse the target sentences in full.

... A semantic effect Christensen () also found increased activation for positive polarity. Compared to negative sentences, positive sentences increased activation in the right supramarginal gyrus (SMG / BA ) in the junction between the right temporal and parietal lobes (Figure ., right panel). This contrasts with Carpenter et al.’s () study, which reported negation to engage BA . Tettamanti et al. () did not find an effect of positive polarity in this region. However, Tettamanti et al. () argued that negation reduces access to the negated concepts by lowering activation in the cortical networks involved. The result from Christensen’s () study is arguably also compatible with this idea. The SMG is part of the multimodal association convergence region involved in lexical retrieval (Fuster ) and in integrating individual concepts into larger concepts (Binder et al. ). According to Russ et al. (), the bilateral SMG, though the effect is larger in the right hemisphere, is specifically involved in the semantic retrieval of enacted action phrases, that is, phrases that are learnt by acting them out rather than by reading them. Positive polarity is upwards entailing, from subset to superset, and as such potentially activates or primes larger semantic networks or representations than negation does. So, negation reduces access to the superset of the negated concept by reducing activation in the semantic convergence zone crucially involved in conceptual representation. Finally, it is also well known that lesions in this general region cause semantic deficits (Damasio ; Pratt and Whitaker ).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

    Left: Neg > Pos

Medial: Pos > Neg (Christensen 2009)



Right: Pos > Neg

 .. The left panel shows where there is more activation for negative sentences (Neg) than for positive ones (Pos), namely in the left BA . Middle and right: The increased activation for positive sentences compared to negative ones in the anterior and posterior cingulate cortex, and in the right SMG, respectively (Christensen : –).

... Polarity and the default mode In addition to the right SMG, Christensen () also reports that the frontal and posterior parts of the cingulate gyrus along the medial surface of the brain bilaterally is more active in positive polarity than in negative polarity (positive > negative), as shown in Figure . (middle panel). In other words, negation decreases activation in the cingulate cortex. The cingulate gyrus is part of the so-called default mode network, a network that has a default (baseline) level of activation, reflecting an intrinsic default brain state (Raichle et al. ; Raichle ; Vanhaudenhuyse et al. ); this bilateral network includes anterior and posterior parts of the cingulate cortex on the medial surface of the brain (i.e. between the two hemispheres) and the junctions between the temporal and parietal lobes. The default level of brain activation is temporarily suspended during attention-demanding tasks resulting in local deactivations. That negation showed decreased activation in the cingulate cortex supports the hypothesis that positive polarity is default, and that a negative statement is initially interpreted as, and derived from, a positive one by adding a negative operator, in the syntax represented as a NegP. Negation is more attention-demanding, more difficult (requires more processing time and induces more errors), which further reduces the default-mode activation.

... Negation and clausal embedding Bahlmann et al. () studied the fMRI effect of negation within and across clause boundaries in German using sentences controlled for length, as in (), where the polarities of the matrix clause and the embedded clause were manipulated independently. () Es ist schon/nicht wahr, It is indeed/not true dass Peter Thomas letzte Woche für das Projekt wirklich/nicht einstellte. that Peter Thomas last week for the project really/not hired. ‘It is indeed/not true that Peter did indeed/not hire Thomas for the project.’

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ø 

Left: Long > Short movement

Left & Right: Neg > Pos

(Christensen et al. 2013)

(Bahlmann et al. 2011)

 .. The left panel: The effect of phrasal movement out of an embedded clause to the front of the matrix clause (Long movement) compared to movement to the front of the embedded clause (Short movement) (Christensen, Kizach, and Nyvad : ). Right panel: Increased activation with matrix clause negation (Neg) scoping into an embedded clause compared to positive polarity (Bahlmann et al. : ).

The sentences were presented incrementally phrase by phrase. Each sentence was followed by a probe question which the participants (n¼) were to judge as true or false (ensuring that they had parsed the sentences in full). The results showed that negation in the matrix clause, but not in the embedded clause, increased activation in Broca’s area, left BA , bilateral inferior parietal cortex, and right insula (Figure ., right panel). The results from this experiment are notably different from the previous ones, apart from the activation in BA . Bahlmann et al. () argue that some of the discrepancy, in particular in the involvement of Broca’s area, might be due to differences in which aspect of sentential negation was studied. They claim that they studied semantic aspects of negation, whereas Tettamanti et al. () studied syntactic aspects, and Carpenter et al. () and Christensen () studied pragmatic aspects. However, Bahlmann et al. () do not provide any justification for this claim, and it is also very difficult to see how negation is examined differently when comparing the examples in ()–() above. Sentential negation has semantic (reversal of truth value and entailment) and syntactic consequences (structure) in all of them, and none of them addresses specific pragmatic aspects of negation (see e.g. () and ()). That being said, it is also clear that Bahlmann et al.’s () study investigates variation in the scope of negation, which has not been investigated with fMRI before. Notably, they found that matrix clause negation increased cortical activation compared to positive polarity, but that it was not the case for embedded clauses —that is, only when the scope of negation crossed a clause boundary into the embedded clause. Many studies have found that moving a constituent to the front of the sentence, as in a whquestion, increases activation in Broca’s area and (sometimes) in the posterior temporal cortex (Wernicke’s area) (Friederici ). Christensen, Kizach, and Nyvad () investigated the effects of wh-movement (in Danish) to the front of an embedded question, compare () and (), and out of the embedded clause to the front of the matrix clause, as in (). ()

Does she know you can buy books there?

()

Does she know what you can buy ___ there?

()

What does she know that you can buy ___ there?

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



The results showed that movement out of the embedded clause increased activation in the left hemisphere, including Broca’s area and BA  (Figure ., left panel), whereas there was no effect of movement to the front of the embedded clause—that is, only when the syntactic movement crossed a clause boundary. The effect in BA  is presumably due to the increased syntactic complexity (due to the word order change), whereas the effect in Broca’s area may be due to the mapping between the matrix and the embedded clause. The same may be the case for Bahlmann et al.’s () results. The matrix clause negation takes scope into the embedded clause, triggering activation in Broca’s area. The projection of NegP in the main clause triggers activation in BA . In both studies, it is conceivable that the processing of the embedded clause masks the effects of embedded negation or movement. One might also speculate that because negation occurs later in the sentence in an embedded clause, there is not enough time for its effects to influence the cortical activation pattern. However, it seems unlikely since the statistical model in the study by Bahlmann et al. () (as is standard) included so-called time derivatives taking such potential delays into account when mapping possible correlations between cognitive task and changes in brain activation. The same applies to the study on wh-movement by Christensen, Kizach, and Nyvad ().

.. N: N  

.................................................................................................................................. The overall pattern here, with negation (and polarity in general) increasing activation in areas outside (and not in) Broca’s area, is also compatible with the data from aphasia studies. Although the relationship between Broca’s aphasia or agrammatism on the one hand and lesions to Broca’s area on the other is not straightforward, there is a clear association between the two (Damasio ; Pratt and Whitaker ). Agrammatism is associated with severe comprehension and productions problems with, among other things, finiteness and verb movement, wh-movement, and clausal embedding (Bastiaanse, Bouma, and Post ; Friedmann ; Friedmann ; Grodzinsky ; Nyvad, Christensen, and Vikner ). There are relatively few published studies on negation and aphasia, and the evidence suggests that negation is relatively spared, or at least not selectively impaired (Fyndanis et al. ; Hagiwara ; Lonzi and Luzzatti ; Rispens, Bastiaanse, and van Zonneveld ). Crucially, there is no comprehension difference between negative and positive polarity, and the production problems reported in these studies seem to have more to do with verb movement than negation. Lonzi and Luzzatti () report that when the French and Italian agrammatic speakers (n¼) in their study produced finite verbs, the verbs correctly preceded negation, whereas when they made mistakes and produced nonfinite verbs instead, these verbs correctly followed negation. Here, negation is clearly intact, and the placement of the verb relative to negation (not vice versa) depends on finiteness. Rispens, Bastiaanse, and van Zonneveld () argue that negation is more impaired in languages where verb movement and negation interact. In their study, the two agrammatic speakers of English made more errors than the three Dutch and two Norwegian agrammatic speakers. Dutch and Norwegian are V languages, which means that in main clauses,

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



 ø 

the finite (tense-inflected) verb is in the second position, it moves to C (see the structure in ()), irrespective of the polarity of the clause. English on the other hand, is not V, and sentential negation with not triggers do-support (compare She likes him, *She not likes him, She does not like him), and if negation is a clitic, it moves with the verb under inversion, for instance in questions (Doesn’t she like him?). Rispens et al. () argue that when the agrammatic speakers made mistakes, they produced constituent negation instead of sentential negation. The production test consisted of two anagram tests in which the participants had to produce a sentence using a set of cards with words on them. Often, the speakers produced sentences without negation, which simply reduces the complexity of the sentence. When negation was included, agrammatic English speakers, for example, erroneously placed the main verb before negation, not after it, resulting in what looks like constituent negation (target: The boy does not congratulate the girl, produced: The boy does congratulate not the girl). Similar results are reported for two Greek agrammatic speakers by Fyndanis et al. (). However, given the bulk of evidence that agrammatic speakers have severe problems with verb movement, word order variation, and structural complexity, it seems likely that the observed behavior is related to the same more general deficit rather than to negation, which is actually produced. This is also supported by the fact that there is no significant difference between positive and negative sentences in comprehension.

.. S

.................................................................................................................................. The linguistic difference between positive and negative sentences has semantic as well as syntactic consequences which also have measurable psycho- and neurolinguistic effects. Psycholinguistic studies have shown that negation is more difficult to process than positive polarity, resulting in increased processing time and increased error rates. The data suggest that a negative sentence is initially processed as positive, then as negative. This is compatible with the (more or less) standard assumption in generative syntax that sentential negation is realized as a negation phrase (NegP), which extends the clause structure and thereby makes it more complex. Semantically, negation inverts the truth values of a statement as well as the direction of the entailment, from upwards entailment (from the more specific to the more general) in positive polarity to downwards entailment. The neurophysiological effects of the interaction between the syntax and semantics of negation have been measured with ERP. NPIs require the presence of an accessible licensor, which means that the NPI must be c-commanded by a negation. Failure to properly license an NPI triggers an N due to semantic integration problems. Intervening (present but not c-commanding) negation affects the amplitude of the N, but does not remove it. Failure to properly license a PPI triggers a P due to syntactic reanalysis and/or potentially additional discourse processing. These results also show that negation is not only a semantic phenomenon, but crucially depends on syntactic structure. Despite variable and sometimes conflicting results, fMRI studies provide a somewhat consistent picture of how negation is processed in the brain. On the semantic side, studies have shown that negation reduces access to negated mental representations by decreasing activation in cortical motor representations for action verbs, and in the temporo-parietal

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

   



junction, more specifically in the right supramarginal gyrus, which is crucially involved in integrating conceptual information and lexical retrieval. The syntactic aspect of sentential negation increases activation in premotor cortex (BA ). One study has found negation to also engage Broca’s area, but this effect is arguably due to the processing of sentential embedding rather than negation. The imaging data also supports the analysis of positive polarity as default, as negation decreases activation in anterior and posterior regions in the cingulate cortex which are part of the default mode network of the brain. Again, this is in line with the assumption that a negative sentence is derived from a positive one, syntactically by adding more structure and semantically by inverting truth value and entailment, and by reducing access to semantic representations. The finding that Broca’s area does not seem to be crucially involved is also compatible with the data from studies on agrammatism, which show that negation is relatively unimpaired in Broca’s aphasia. Negation is central to human language, and understanding how it interacts with cognition and how it is processed in the brain is of great interest to the science(s) of language. Though the emerging picture of how negation is processed in the brain is still showing mixed and conflicting results, this variation is most likely due to methodological differences. The overall picture is consistent with converging data from linguistics (syntax and semantics), psychology (processing time and error rates), neuroimaging (fMRI), neurophysiology (ERP), and from aphasiology (agrammatism). However, further studies are needed to (dis)confirm it, studies that replicate the experiments discussed here, while controlling for the various confounds (scanning only parts of the brain, using passive listening, using stimuli with non-equal length). It would also be interesting to combine electrophysiological and fMRI techniques in one and the same study in order to study the time course as well as the localization of negation in the brain. Another theoretically interesting avenue would be to perform parallel fMRI studies in two (or more) languages to see if there are any systematic cross-linguistic differences in how the brain processes negation which might, for instance, be related to how negation is realized (say, as an adverb or as an affix). Finally, the possible effect of brain damage on negation is still relatively understudied.

A

.................................................................................................................................. The author would like to thank Anne Mette Nyvad, Sten Vikner, and two anonymous reviewers for helpful comments on an earlier version of this chapter.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

  ......................................................................................................................

                   ......................................................................................................................

     

.. I

.................................................................................................................................. I less than two decades, the neurobiology of language has seen tremendous progress, thanks to increasingly available neuroimaging methods to study the human brain function, and an effort for integrating results from brain research with knowledge from cognitive psychology, computer science, psycholinguistics, linguistics, and philosophy. An important part of this work investigates how the meaning of words, sentences, and discourses is represented in the brain; that is, how humans understand language (e.g. Binder and Desai ; Flick et al. ; Friederici ; Hagoort et al. ; Kemmerer ; Kutas and Federmeier ; Mason and Just ; Papeo and Hochmann ). Current models, conceived to account for the cortical representation of meanings (e.g. the meaning of the word speaking), hardly accommodate the representation of negated meanings (e.g. not speaking). Research on negation appears particularly scant, compared with the bulk of research on other aspects of language. And it is so, despite the general agreement on the universality of negation in human language and thought and its uniqueness in human cognition (Horn ). The state of affairs in the neurobiology of language might reflect the lack of consensus among philosophers, linguists, psycholinguists, and cognitive scientists on the definition of negation (Horn ; Givón ), the underlying cognitive abilities and processing stages (Clark ; Fodor, Fodor, and Garrett  Wason and Jones ; Wason ), and the developmental trajectory of logical negation (Austin et al. ; Pea b). Despite the difficulty, research on the neural mechanisms of negation in the context of word or sentence understanding has reached a critical mass. It can thus be asked whether the study of the human brain has taught us something about the mechanisms for understanding negation and this will be the focus of this chapter. The first part of the chapter briefly reviews the traditional models of processing negation, derived from behavioral experiments on sentence comprehension and verification.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

      



Traditionally, these models offer variants of a two-step view of sentential negation processing, where an initial default representation virtually identical to that of the affirmative (counterfactual) statement is followed by an additional processing that yields a realistic representation of the actual situation. These two-step functional models have been relevant to guide research on the neurobiology of negation. The second part of this chapter focuses on current literature that has used various methodologies to measure human behavior in cognitive tasks, and its neural correlates, such as the changes in the activity of brain areas recorded with functional magnetic resonance imaging (fMRI), the changes in the electrical activity of neuronal populations recorded with electroencephalography (EEG) and the changes in peripheral muscles induced by non-invasive brain stimulation combined with electromyography (EMG). This corpus of studies has measured the behavioral or neural responses to full negative and affirmative sentences (e.g. With proper equipment, scuba-diving isn’t very dangerous; example from Nieuwland and Kuperberg ) or to processes that depend on the understanding of those sentences (e.g. during processing a picture for sentence-picture verification). A subset of those studies has tried to isolate the neural response to the argument in the scope of negation (e.g. the response to a given word, when preceded by negation, such as non scrivo, I don’t write; example from Papeo, Hochmann, and Battelli ). Across those studies, a consistent—although not ubiquitous—finding is that negative phrases or sentences yield reduced levels of activity in regions involved in the representation of the corresponding affirmative meanings. In the light of such neural effect of negation, two issues have been investigated. First, when does the change in the neural activity associated with negated words or sentences occur? Second, what mechanism modulates the level of activity in a brain network, thus to mark the distinction between a meaning (e.g. eating) and its negative counterpart (not eating)? The third part of the chapter addresses these issues. This last part unfolds in two main strands of discussion. First, the time course of neural response to negated words or sentences is discussed to characterize the different temporal dynamics, that is the moment-by-moment changes in the brain activity, during the course of processing negative versus affirmative sentences. This literature has singled out the popular view that, relative to affirmation, negation entails an extra processing stage, which makes it inherently slower and more complex. This view is implied in traditional (Russell ; Carpenter and Just ; Just and Carpenter ) as well as more recent models of negation (Hasson and Glucksberg ; Kaup et al. ). It is considered whether and to what extent the current analysis on behavioral and neural correlates of negation supports this view. Second, the last part of the chapter discusses the possible neural mechanisms that can operate to mark the representation of negative versus affirmative meanings. Based on recent research, it proposes the hypothesis that processing negated meanings involves two functionally independent networks: the response inhibition network and the lexicalsemantic network representing the word/words in the scope of negation. The link between negation and inhibition is recurrent in the debate on the ontogenesis of negation, where various perspectives have brought up the “prohibitive” or “inhibiting” connotation of negation (Clark ; Pea b; Russell ; Vandamme ; Wason and Jones ). However, only recently have researchers in cognitive neuroscience begun addressing this link.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



     

This chapter aims at providing an overview of the literature that has forwarded a certain interpretation of negation (negation is harder than affirmation and requires an extraprocessing stage) or that has questioned that (traditional) interpretation. The chapter intends to offer elemental information to evaluate the body of evidence in favor of one or the other model of negation. Moreover, the discussion in this chapter does not exhaust the multiple processes associated with negation. It seeks to characterize the known effects of negation on the neural representation of word/sentence meanings and to identify the brain network, or networks, that can be recruited during processing negation. In doing so, it lays the foundations for a biologically plausible model of processing negated word meanings.

.. M         

.................................................................................................................................. A traditional model holds that for negative sentences, the affirmative meaning is first fully activated and then modified, or receded, to reflect negation (Carpenter and Just ; Chase and Clark ; Russell ). The sentence verification task has been frequently used to address this so-called modal processing model of negation. In such a task, a sentence is first presented (e.g. It is not true that the dash is above the plus) and then followed by a picture (±) congruent or incongruent with the sentence. The task for the participant is to decide whether the sentence is true or false. Slower response times have typically been reported, when the sentence involved negation of a predicate (It is true that clouds aren’t heavy) or of a full sentence (It isn’t true that clouds are heavy). In Carpenter and Just’s account (), a sentence is first represented as an affirmation and then re-scanned to verify the presence of negation; the presence of a negation marker would trigger the receding process to identify the constituents to recede (see also Chase and Clark ). The specific aspect of the modal processing model, that is, the involvement of a process for the comparison of constituents (or receding process), has been primarily studied with the sentence verification task. Other paradigms have been used to address the questions as to whether negation impacts (i.e. delays) task performance (e.g. a lexical-decision task and a phoneme-monitoring task; see Cutler ), or whether it reduces the level of activation of a concept (e.g. probe task; see MacDonald and Just ). The more general stance that processing sentential negation entails two steps, common to the modal processing model and the two-step simulation model of negation, has been investigated with various methodologies, as we review in the following. The modal processing model shared this stance with the two-step simulation model of negation. The two-step simulation model of negation has resumed the view that negation entails an additional process (Kaup, Lüdtke, and Zwaan ; Kaup et al. ). On this account, understanding a negative sentence (e.g. The elephant is not above the giraffe) involves a first step in which the comprehender generates a representation of the counterfactual situation (i.e. a mental image of an elephant above a giraffe) and, later, a representation of the actual

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

      



state of affairs (i.e. a mental image of an elephant below a giraffe). The processing of negation would lie in the matching of the two representations. In support of this model, it has been shown that a negative proposition (e.g. There was no eagle in the sky) facilitates the processing of an image representing the counterfactual situation, presented immediately after the sentence (e.g. an eagle with outstretched wings, as if flying in the sky; see Kaup et al. ). The two models of negation differ in many respects. An important difference concerns the number of representations triggered by a negative utterance. In the modal processing model, only the affirmative representation is generated, and a revision of this representation is achieved through the receding process. In the two-step simulation model, two models, one for the state of affairs (negated representation) and one for the corresponding affirmative situation, are generated and then matched. Despite this difference, both models predict that verification times are longer for negative than for affirmative sentences, reflecting the additional processing stage for receding or matching. We note that this conclusion, mostly based on research on sentential negation, has been sometimes generalized to lexical negation and other semantic structures (see Carpenter and Just ; but see Cutler ). However, that experimental research on lexical negation appears underdeveloped, relative to the study of sentential negation. Another difference between the two models concerns the format of representation, which is thought to be propositional in the former case, and a mental model or simulation in the latter. Addressing the discussion on the format of representation is outside the scope of this chapter (for a discussion see Machery ). However, the idea that language understanding entails non-propositional representation of the sensory and/or motor properties of the word’s referent as experienced in the real world, has heavily impacted the research on negation. In particular, a corpus of studies has documented activity in motor brain areas, with a resonance at the level of peripheral muscles, during understanding language describing motor actions (e.g. kicking) (Papeo, Rumiati, et al. ; Tettamanti et al. ; Pulvermuller ). Since the correlates of motor cortex activity are relatively easy to access by various methodological approaches, many studies on negation have measured those correlates to establish how the neural representation of word or sentence meanings changes with negation. Thus, the current literature on the neural processing of negation might suggest a general agreement on the involvement and role of sensory-motor representations in lexical-semantic processes. However, this is not the case, as many controversies remain in the field. In particular, although motor activity is consistently observed in the earliest stages of processing action-related words (within  ms from the word onset), it appears secondary and contingent to activity in the association cortices involved in lexical-semantic processing (e.g. the middle temporal gyrus; Papeo et al. ), and its function for language understanding remains unclear (Kemmerer ; Papeo and Hochmann ). With this clarification in mind, the upcoming sections will review recent research, where languagerelated motor activity has been used as a proxy for the activity in the larger frontotemporal-parietal network for lexical-semantic processing. Before proceeding, we shall clarify that, particularly in the context of fMRI research, the term “increased/reduced activity” refers to the relative change in the amplitude of neural activation in a given brain region (i.e. the fluctuation in the blood oxygenation level dependent—BOLD—signal).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



     

.. D       

.................................................................................................................................. With some exceptions (Giora et al. ), behavioral studies have consistently shown different verification times for affirmative and negative sentences (Carpenter and Just ; Chase and Clark ; Kaup, Lüdtke, and Zwaan ; Kaup et al. ; MacDonald and Just ). This circumstance has maintained that the processing of affirmative and negative sentences differs and that longer processing times for negative sentences reflect greater computational complexity and cognitive demand. Based on this idea, early neuroimaging research has used negative sentences to demonstrate that the levels of activity in task-relevant brain regions, as measured with functional magnetic resonance imaging (fMRI), varied with task complexity. Carpenter and colleagues (Carpenter et al. ) tested eight individuals on a sentence-verification task during fMRI. They reported that, relative to the condition with affirmative sentences, in the negative sentence condition (e.g. It is/isn’t true that the star is above the plus) participants took longer to read the sentence, process the picture that followed the sentence, and match the two for sentence verification. Moreover, increased processing times for negative sentences were associated with increased activity in regions of interest as defined by anatomical landmarks: the posterior superior/middle temporal lobe and the supramarginal and angular gyri in the parietal lobe. More recently, Christensen () used fMRI during a probe-to-target matching task. In that task, participants saw a sentence for five seconds (e.g. the toughest men know but don’t use soap) followed by another sentence (e.g. the toughest men use soap); they had to decide whether the latter was true or false, given the former. Like Carpenter et al. (), Christensen found that, relative to affirmative sentences, negation increased the processing times regardless of whether it was in the probe, or in the target preceding the probe. Moreover, negative sentences increased neural activity. However, this effect occurred in a different site (i.e. the left precentral gyrus), than in Carpenter et al. (), and other brain areas showed the opposite effect (higher activity for affirmative versus negative sentences was found in right supramarginal gyrus and anterior and posterior cingulate cortex). Those attempts to study negation in the brain converged with the above-mentioned behavioral results in suggesting that negation differs from, and might be more difficult than, affirmation; as such, it would recruit larger neural resources. However, those studies have left unanswered the question as to how understanding negation differs from understanding affirmative utterances. A way to approach this question would be to isolate the neural activity specifically associated with the representation of word meanings from neural activity reflecting task difficulty and cognitive demand in general, and to study how the former changes after negation. A set of studies has undertaken this approach, targeting language-related motor activity. As anticipated above (see section .), motor activity has been consistently associated with the retrieval of word meanings with motor content (Kemmerer ; Papeo, Pascual-Leone, and Caramazza ; Reilly et al. ). In favor of a link between motor activity and

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

      



semantic processing of words (or representation of word meanings), it has been shown that language-related motor activity is task-specific, that is, it is maximal during the semantic analysis of words vs. other types of tasks (Papeo et al. ; Papeo, Rumiati, et al. ). Moreover, language-related motor activity is category-specific. That is, it is specific to words/sentences with motor-action contents, and does not generalize to other contents, even when the linguistic material had been matched for variables (word frequency, length, and so on) that could determine an unbalanced recruitment of processes and resources (e.g. Kemmerer ; Reilly et al. ). Finally, decoding of information using multivariate analyses of neural data shows that language-related motor activity does carry information about action contents described by language (Wurm and Caramazza ). As such, as limited as it might be, motor activity provides a window onto the larger network associated with word semantics (e.g. Binder and Desai ; Lambon-Ralph ). Using electroencephalography (EEG), Alemanno et al. () measured the event-related desynchronization (ERD) of mu-rhythm, a correlate of motor and premotor cortical activation involved in movement preparation and execution (Pfurtscheller and Lopes da Silva ). Mu-rhythm ERD was recorded while Italian participants read silently affirmative or negative sentences describing hand-related actions (Io scrivo, I write, Non scrivo, I don’t write) or non-motor contents (Io penso, I think, Non penso, I don’t think). The results showed: () greater mu-rhythm ERD for hand-action phrases than for abstract phrases over the left premotor and motor areas, and () delayed and increased mu-rhythm ERD for negative than for affirmative action-related sentences. The authors interpreted the former result as reflecting a greater involvement of motor regions in processing actionrelated language as compared with non-motor contents, and the latter result as reflecting a greater difficulty of processing negative as compared with affirmative phrases. Although one might be concerned about two different interpretations of the same physiological effect (i.e. increased mu-rhythm ERD reflecting task-relevant activity for hand-action versus non-motor verbs, and task difficulty for negative versus affirmative phrases¹), Alemanno et al.’s results agreed with the traditional view that negation is more difficult than affirmation (for EEG effects of negation in Turkish, see Yanilmaz and Drury ). However, how negation differed from affirmation was yet left unknown. Two fMRI studies have provided initial evidence that, besides difficulty/complexity differences, negation operates by reducing the activation of, and/or the access to the representations of the meaning in its scope. Tettamanti and colleagues () asked participants in the fMRI scanner to listen to affirmative and negative sentences with motor-action or non-motor content. Examples of sentences are: Adesso io premo il bottone, Now I push the button (affirmative motor-action sentence); Adesso non premo il bottone, Now I don’t push the button (negative motoraction sentence, with dropped pronoun); Ora io apprezzo la fedeltà, Now I appreciate the loyalty (affirmative non-motor sentence); and Ora non apprezzo la fedeltà, Now I don’t appreciate the loyalty (negative non-motor sentence, with dropped pronoun). The authors ¹ Since Italian is a pronoun-drop language, the authors decided to omit the pronoun ‘io’ (I) in the negative sentences to keep constant the number of words in the affirmative- and negative-phrase conditions. In effect, as the authors themselves acknowledged, the absence of a pronoun in the negative condition, as opposed to the negation itself, could account on its own for the increased difficulty, in that extra effort may be required to retrieve an implied pronoun.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



     

reported that, overall, negation deactivated cortical and subcortical structures (right middle frontal and middle occipital gyri, and left pallidum). Moreover, having identified the brain network that responded more strongly to motor-action than non-motor sentences, they found that negation reduced the activity in a large part of that network, encompassing left inferior frontal, inferior parietal, and middle temporal areas, and reduced the functional integration (i.e. the coupling of activity) within that network. Tomasino, Weiss, and Fink () identified the individual’s hand representation in the primary motor and premotor cortices, during performance of simple hand movements by participants in the fMRI scanner. Then, they measured the neural activity within those regions of interest while participants read silently affirmative and negative hand actionrelated sentences in English (Do grasp, Don’t grasp), interleaved with pseudo- sentences (i.e. phrases formed with pseudowords, e.g. Do gralp, Don’t gralp) for a lexical decision task (i.e. deciding whether the word existed or not). Lexical decision was slower with negative than affirmative sentences, and neural activity in the bilateral primary motor and premotor cortices was lower for negative than for affirmative hand-action sentences. While Tettamanti et al. () endorsed an account of the neural effects of negation in terms of reduced activation/accessibility of information, they anticipated—as did Tomasino, Weiss, and Fink ()—the possibility that negation entails inhibition of processes or responses associated with word understanding. Before asking how negation can be characterized in terms of neural processes, a preliminary question may be: if the neural representations of affirmative and negative utterances differ, when do they differ? As discussed in the following, this question crosses the debate around the view that representation of negation requires a first step, in which the counterfactual, affirmative representation is retrieved.

.. T    : O   ?

.................................................................................................................................. Traditional studies on negation have shown that verification times are slower for negative sentences than for the corresponding affirmations (see section .). Other studies have used paradigms based on semantic priming to characterize the time course of processing negation. For example, Kaup, Lüdtke, and Zwaan () presented a sentence (e.g. The umbrella was not open) followed by an image that could match (a closed umbrella) or mismatch (an open umbrella) the sentence. Participants, asked to name the object in the image, were faster in the match- than in the mismatch-condition, but only when the image was presented  ms after the sentence. With a shorter delay ( ms), naming time was comparable for matching and mismatching images, suggesting a more ambiguous representation of the sentences—between the negative and the affirmative version—at early stages of negation processing. The authors suggested that the negated meaning becomes fully available only after an initial stage in which the corresponding affirmative meaning is also activated (Kaup, Lüdtke, and Zwaan ). Hasson and Glucksberg () reached similar conclusions using affirmative and negative metaphoric sentences (e.g. These lectures are/are not sleeping pills) followed by a

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

      



target-word, which could be related or unrelated (e.g. interesting or tedious) to the sentence meaning. Naming the target-word related to the affirmative meaning (tedious) was facilitated by both negative and affirmative sentences, when the target-word appeared  or  ms after the sentence offset. Naming the word related to the negative meaning (interesting) was facilitated by negative sentences only when presented  ms after the sentence offset. The authors concluded that the negative representation was available only in a secondary stage, following the activation of the affirmative representation. After these sorts of results, new research has questioned the linguistic context in which negation was tested, and the temporal scale afforded by reaction times in explicit tasks such as sentence verification and lexical/semantic decision. This research is addressed in the following.

... Interaction between pragmatics and negation A task that requires matching a sentence with a picture can solicit the formation of a mental visual image. While an affirmation defines a given state of affairs, and can therefore elicit a specific image, a negative statement such as There is no eagle in the sky (example is from Kaup et al. ) implies a virtually infinite set of possibilities (e.g. if there is no eagle in the sky, there could be a star, a plane, a pigeon, a butterfly, etc., and the eagle could be in the nest, on a tree, diving in water, and so on and on). In the latter case, the affirmative, counterfactual state of affairs (an eagle in the sky) could be represented because it is the only specified representation. Such a mechanism would lead to faster recognition of an image depicting the affirmative state of affair (an eagle in the sky), even if the individual has understood the negation. Orenes, Beltrán, and Santamaría () illustrated this point using the visual word paradigm (Altmann ; Cooper ; Tanenhaus et al. ). It is known that individuals automatically move the eyes toward the visual input, if available, that matches a concurrent verbal message. Orenes and colleagues showed that, with only two available alternatives (e.g. a red and a green figure), participants promptly represented the alternative of a negative statement (i.e. green for the sentence: The figure is not red). However, when an obvious alternative was not available (i.e. in the context of multiple alternatives: red, green, yellow, or blue), they represented the negated argument. Thus, the activation of the affirmative representation would not be mandatory and could depend on the number of possibilities opened by a negation. Other pragmatic aspects of the task have proven crucial to driving the processing of negation. In everyday exchanges, negation is used in reference to an earlier situation or statement that is rejected. In most experiments, negation has been tested with simple sentences (e.g. I don’t write, The umbrella is not open, There is no eagle in the sky) presented without a context, which could make the message appear awkward, trivial, or uninformative (Grice ; Sperber and Wilson ). Nieuwland and Kuperberg () used EEG to measure the N, an event-related potential (ERP) component that indexes the semantic integration of a word with the preceding linguistic context (Kutas and Hillyard ). They showed that negative sentences were processed just like affirmative sentences when they were licensed, namely, when the negated word meaning was informative (e.g. With proper equipment, scuba-diving isn’t very dangerous), but not when negation was uninformative

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



     

(e.g. Bulletproof vests aren’t very dangerous). The use of unlicensed negation could account for the results of studies that, using a similar methodology, found that the integration of negation into a sentence meaning took place only at late latencies ( ms) (Lüdtke et al. ). Further studies (Nordmeyer and Frank ; Orenes et al. ; Tian, Breheny, and Ferguson ; Zhang and Pylkkänen ) have emphasized the interaction between pragmatics and negation, questioning the necessity of an additional processing step, which would make negation inherently harder than affirmation.

... The default computation of negated meanings One might argue that, even if certain pragmatic contexts may favor an incremental interpretation of negation, a model with two stages illustrates the default way in which the argument in the scope of negation is processed. That is, a model where negation operates only after the activation of an affirmative (counterfactual) representation might account for the computation of negation, unless specific pragmatic elements constrain the interpretation of negated word meanings. Such a model predicts that, at least in the initial stage of semantic retrieval, the neural representation activated for a given word (e.g. writing) and for its negative counterpart (not writing) would be indistinguishable. Testing this prediction would require tracking semantically related activity throughout reading (or listening to) a word versus a negated word, before other integrative processes begin. A number of studies have characterized neural activity at late stages, when semantic information is integrated at the sentence level with pragmatic information, or during sentence verification (Liuzza, Candidi, and Aglioti ; Fischler et al. ; Nieuwland and Kuperberg ). An exception is the study by Lüdtke et al. (), who reported EEG activity during reading affirmative and negative sentences (e.g. Vor dem Turm ist ein Geist, In front of the tower there is a ghost; Vor dem Turm ist kein Geist, In front of the tower there is no ghost), for a sentence-picture verification task. Interestingly, they showed a change in neural activity as soon as the marker of negation (kein, no in German) was encountered; but, as the authors noted, the topography and time course of the effect could not be unambiguously ascribed to semantic processing. Thus, the earliest effects of negation, on the representation of a word meaning, remained unclear. Papeo, Hochmann, and Battelli () have addressed this issue testing participants in a pragmatically neutral context, during silent reading of affirmative and negative sentences (ora scrivo, Now I write; or non scrivo, I don’t write). Using transcranial magnetic stimulation (TMS), the authors targeted the motor portion of the left precentral gyrus that is recruited by the lexical-semantic brain network, in the earliest stage of processing actionrelated words (within  ms from the word onset; Papeo et al. ; Papeo and Caramazza ; Shtyrov et al. ). A TMS pulse delivered over the motor precentral cortex induces a visible muscle twitch in the peripheral muscles controlled by the stimulated area (e.g. right hand muscles for TMS on the hand-area of the left motor precentral gyrus). The amplitude of the twitch, recorded in the form of motor-evoked potentials (MEPs), provided a measure of motor activity (i.e. motor excitability) at the exact time of TMS delivery: , , or  ms after the word onset (a first-person action-related verb—e.g. scrivo, I write— preceded by the negation non, don’t, or by the adverb ora, now). Sentences matched for

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

      



frequency and length but with non-motor content (e.g. ora/non penso, Now I wonder/I don’t wonder) were included to verify that the target-measure (MEPs) isolated the effect of word semantics. The authors reasoned that, if a negated word meaning is initially represented like a counterfactual affirmation, motor excitability should be comparable for both the affirmative and negative condition, at least in the early stage of semantic processing (~ ms). Early differences between the affirmative and negative condition would be in keeping with the view that negation is incrementally incorporated to construct a meaning. The results, shown in Figure ., supported the latter conclusion. Motor excitability increased selectively for motor-action words in the affirmative condition, while it was significantly lower for the same words presented after negation. In the latter condition, excitability did not differ from the activity associated with non-motor words. Importantly, the difference between affirmative and negative motor-action words was found at the earliest probed timing ( ms), despite the impoverished sentential context where negation was not pragmatically licensed (see Figure .). Papeo et al.’s results showed early effects of negation in the motor areas associated with processing motor-action words. Likewise, the fMRI study by Tettamanti et al. (; see (A)

0.71 0.69

MEPs (mV)

0.67 0.65 0.63 0.61 0.59 Positive

0.57

MEP (mV)

(B)

0.71 0.69 0.67 0.65 0.63 0.61 0.59 0.57

250

400

Negative

550 (ms)

250

action verbs

400

550 (ms)

state verbs

 .. Results of Experiment  in Papeo et al. (). A. Mean amplitude (mV) of the motor evoked potentials (MEPs) induced in right-hand muscles by TMS delivery over the hand-representation in the left motor precentral gyrus, during reading of hand-action and non-action verbs in the affirmative and in the negative context, averaged over time and B. as a function of time intervals. MEP amplitude, as a measure of corticospinal excitability, was overall higher for action verbs presented in the affirmative context (preceded by ora, now) than in the negative context (preceded by non, don’t), in the time interval corresponding to lexical-semantic access (– ms after the verb onset).). No difference was found for non-action (state) verbs in the negative vs. afformative context. Error bars denote within-subject standard errors.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi



     

section .) showed decreased neural activity in frontal, parietal, and temporal sites associated with action-related language. The latter study, however, provided no information on the time course of activity. Thus, it remains an open question whether the temporal course described by Papeo Hochmann, and Battelli () for word-related motor activity applies to the extended network representing word meanings. The literature reviewed here suggests that negation reduces, blocks, or inhibits some processes that normally occur during processing of word meanings, such as the activation of motor areas. Multiple terminologies (reduces, blocks, or inhibits) are used purposefully, to highlight the open question concerning the neural mechanism that, by affecting neural activation, could mark negation in the brain. In addressing this question, recent research has emphasized a functional link between inhibition and language processing as a possible mechanism underlying the computation of negation.

.. T     

.................................................................................................................................. Inhibitory processes are ubiquitous in human cognition and performance due to their adaptive role in changing environments and needs. Individuals continuously monitor the current situation to be able to rapidly inhibit actions or suppress alternative behaviors that are inappropriate, unsafe, or no longer required (Chambers, Garavan, and Bellgrove ). Inhibitory mechanisms can also operate to suppress previously activated representations (Nakata et al. ; Smith, Johnstone, and Barry ). In some reflections on the ontogenesis of negation, the logical not would derive from a more basic representation of the meaning no, acquired early in life in association with unpleasant feelings (Russell ), or with the command for inhibiting an ongoing behavior (Vandamme ); in general, with “prohibitive” or “inhibiting” connotation (Wason ; Wason and Jones ; see also Pea b). Indeed, in early manipulative scenarios, adults frequently use negative imperatives (e.g. ‘don’t touch that’) to guide the child’s behavior (Park ; Austin et al. ). In this way, a child could learn the pragmatic function of negative imperatives as stop signals to suppress the initiated or intended actions. The pragmatic function of negative imperatives as stop signals works also in adults’ communication. For instance, if a friend tells you ‘don’t drink that!’ while you are bringing a drink to your lips, probably you will stop doing it immediately. The use of imperative negation to stop or prevent actions could establish connectivity between the brain network for lexical negation and the networks for response inhibition or conflict monitoring. The literature reviewed above has often described the effects of negation as suppression or reduced availability of conceptual representations (Kaup, Lüdtke, and Zwaan ; Kaup et al. ; MacDonald and Just ), and neural information (Papeo, Hochmann, and Battelli ; Tettamanti et al. ; Tomasino, Weiss, and Fink ). Such characterization somehow implies that understanding sentential negation entails the neurophysiological mechanisms of response inhibition. However, inferring the involvement of inhibitory mechanisms from evidence of reduced neural activation following negation is not so straightforward. Recent studies have directly addressed the hypothesis that the

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

      



system for response inhibition in the brain is recruited during processing negation. Using behavioral methods, neurostimulation, or EEG, those studies provide the first pieces of evidence for a role of inhibition in processing negation.

... Measure of activity in inhibitory neurons during processing negation Having reported reduced motor excitability in response to negated action-related words, Papeo, Hochmann, and Battelli () asked whether motor activity was abolished in the context of negation or, instead, it still participated in the process, but through increased inhibitory, as opposed to excitatory, activity. With a variation of the TMS protocol used to measure motor excitability (see section ..), the authors measured the cortical silent period (CSP), a measure of inhibition. In particular, when a muscle is voluntarily contracted, the TMS delivery over the aspect of the motor cortex that controls that muscle induced first a MEP (and a visible muscle twitch), and then a suppression of the EMG activity (see Figure .A). The duration of such suppression (in ms) corresponds to the CSP duration, which correlates positively with the activity of GABAergic inhibitory neurons (Cantello et al. ; Inghilleri et al. ). Measured  ms after the word onset, the inhibitory activity indexed by the CSP duration was significantly lower for action-related words (scrivo, I write) than for the same words preceded by the negation non (don’t), as Figure .B illustrates. Thus, compared with affirmative language, negation changed not only excitatory activity (lower for negative than for affirmative language) but also the inhibitory activity associated with the processing of action-related words (lower for affirmative than for negative language). Both neural effects of negation took place as early as  ms after the word onset, that is presumably during the retrieval of the word meaning, before the beginning of integrative processes. Papeo, Hochmann, and Battelli’s () study is limited to the case of action-related words and to an aspect of brain (the precentral motor cortex) that lays at the periphery of the semantic network. Those results call for an investigation of the effects of negation in other, more central, aspects of the network representing word meanings. Furthermore, the functional link between negation and inhibition requires characterizing the inhibitory mechanism or mechanisms that are recruited in processing negation. This characterization will be the focus of the upcoming sections.

... Measure of neural signatures of inhibition in processing negation A well-established experimental paradigm in the literature of motor inhibition is the GoNogo task, in which the subject has to provide a motor response to a Go cue that appears frequently (e.g. % of trials), and has to refrain from responding to a less frequent Nogo cue (e.g. % of trials). The Nogo trials require suppressing the preponderant motor response induced by the high frequency of Go trials; as such, Nogo trials require inhibitory



     

(mV)

(A)

× 10–4 6 MEP 4 2 0 –2 –4 –6 –8 –200 –100 0

206 - trial 7

100

200

300

400

500 (ms)

173

(B)

172

CSP (ms)

171 170 169 168 167 166

Positive action verbs

Negative state verbs

 .. Result of Experiment  in Papeo et al. (). A. Illustration of the Cortical Silent Period (CSP) from one trial of one subjects (Participant , Trial ). The CSP corresponds to the period of EMG activity suppression that follows the Motor-Evoked Potential (MEP) elicited by the TMS delivery. Time  denotes TMS delivery. The beginning and the end of the CSP as detected by our automated algorithm, are indicated by the two crosses on the EMG signal. B. Mean duration (ms) of the cortical silent period (CSP) during reading hand-action and non-action verbs in the affirmative and in the negative context. The CSP duration, as an index of inhibitory activity, was significantly higher for action verbs presented in the negative context (preceded by non, don’t) than in the affirmative context (preceded by ora, now). No difference was found for non-action (state) verbs in the negative vs. affirmative context. Error bars denote within-subject standard errors.

resources. Recording EEG during the Go-Nogo task has shown robust neural signatures of inhibitory processes. Particularly, event-related synchronization (ERS) of fronto-central theta band (– Hz) rhythms (Nigbur, Ivanova, and Stürmer ; Huster et al. ; Harper et al. ) and enhanced amplitude of the N, N, or P components of the ERPs (Bokura, Yamaguchi, and Kobayashi ; Maguire et al. ) are consistently observed during Nogo trials. De Vega et al. () took advantage of those measures to study the recruitment of inhibition during processing negation. They asked participants to read Spanish handaction sentences with affirmative or negative polarity (i.e. Ahora sí [no] cortarás el pan, Now you will [will not] cut the bread), with an embedded Go-Nogo task to perform over a cue that appeared  ms after the verb onset. As expected, the analysis of the EEG signal provided strong ERS of theta rhythms in Nogo trials, indexing motor inhibition. Crucially, this effect was qualified by an interaction between cue and polarity. Namely, for Nogo trials, negative sentences diminished the fronto-central theta rhythms ERS compared to

      



affirmative sentences, whereas for Go trials no effect of sentence polarity was obtained. Such interaction was interpreted as an indication of shared inhibitory mechanisms during response suppression and linguistic negation processing. In particular, the negated verb, appearing before the cue, would recruit the inhibitory network and reduce the inhibitory resources (i.e. the ERS of theta) available for the response suppression in Nogo trials. In another EEG study, Beltrán, Muñetón-Ayala, and de Vega () used a variant of the above double-task paradigm and analyzed the ERPs. Participants saw affirmative or negative hand-action sentences embedded in a Stop–Signal task (SST). In the typical SST procedure, participants are always presented with a Go signal, but in some trials, after a variable delay, they also received a Stop signal (e.g. a tone) to withhold the already initiated response. The Stop-Signal delay (SSD) contingently varies from trial to trial so as to produce around % successful stops. This so-called staircase procedure allows computing the stop-signal reaction time, namely the time required to successfully suppress an underway response. The results showed that the amplitude of the N component, an early signature of inhibition processes, was affected by an interaction between sentence polarity and performance in stop trials (success vs. failure in withholding the response). In particular, N amplitude was larger for successful trials with negative sentences than for successful trials with affirmative sentences, whereas no polarity effect was found in unsuccessful trials. The source of these modulatory effects of polarity was the right inferior frontal gyrus, a central component of the neural network for response inhibition (Aron, Robbins, and Poldrack ). Converging with electrophysiological effects, reaction times to the stop signal showed that participants were significantly faster at inhibiting responses in the context of affirmative sentences than in the context of negative sentences (see also Garcia-Marco et al. ). The results of the above EEG experiments encourage the thinking that linguistic negation consumes neural resources of response inhibition, modulating both the EEG rhythms in the fronto-central theta rhythms and the N component of the ERP, as well as increasing the cost of inhibition. However, the above experiments involved exclusively hand-action language; thus, it is not clear whether inhibition is generally involved in negation or is rather specific to the phenomena linked to motor response to actionrelated contents. In a recent study, Beltrán et al. () employed the same dual-task paradigm as in de Vega et al. (), namely, a Go/Nogo task embedded in a sentence comprehension task with manipulation of polarity. This time, both action and non-action related sentences were included. As in de Vega et al.’s () study, the time-frequency analysis of the EEG in Nogo trials showed reduced ERS of theta power in the context of negative sentences compared to affirmative sentences. Moreover, the modulation of polarity was statistically similar for both action and non-action related sentences (see Figure .). This circumstance suggests that the recruitment of inhibitory mechanisms is a general characteristic of linguistic negation, likely operating in the comprehension of any sentential negation. In sum, the set of studies discussed in this section provide initial empirical evidence in favor of the hypothesis that the comprehension of sentential negation recruits the neural circuitry of inhibition. This line of research has begun addressing one of the most ancient (and perhaps intuitive) views of negation as grounded in the neural mechanisms for monitoring and inhibition of behaviors and representations.

30 18

FCZ

NG NEGATIVE FCZ

1.5

DIFFERENCE

0.5

(B)

AFF

NEG

FCZ

12

DIFF 1.0

0.3

–1.0

–0.3

6 2

18

GO AFFIRMATIVE FCZ

GO NEGATIVE FCZ

DIFFERENCE

GO

3

30 Frequency in Hz

NG AFFIRMATIVE

FCZ

12

DIFF

Frequency in Hz

(A)

     

NG



6 3 2

–0.8 –0.4

0.0 0.4

Polarity Verb marker GO/NG cue

Time in seconds

–0.8 –0.4

0.0 0.4

GO reaction times distribution

–0.6

Relative Change

0.6

4–8 Hz 210–650 ms

Significant electrodes

 .. Results of Beltrán et al.’s study (submitted). Modulation of the inhibition-related theta rhythms by sentence polarity in a cluster of significant electrodes. Time-frequency analysis of the EEG inhibition-related theta rhythms for the interaction Cue (Go/Nogo) x Polarity (affirmative / negative) averaged for action (you will [will not] cut the bread) and non-action sentences (you will [will not] trust the man). A. The left-side boxes show that affirmative-Nogo trials elicited larger theta power than negative-NoGo trials in a relative early interval associated with response inhibition, whereas no effect of polarity was observed in Go trials. B. Distribution on the scalp of the theta-band modulations in the significant cluster of electrodes.

.. C

.................................................................................................................................. The goal of the neurobiology of language is to describe how the human brain realizes the language function. This goal has appeared particularly difficult in the case of lexical or sentential negation. Yet, the current body of studies using various measures of brain functioning has identified some solid facts. Negated word meanings can be processed within the same temporal interval as their affirmative counterparts. The implication, also compatible with recent behavioral studies, is that negation is not necessarily harder than affirmation. Although processing negation appears particularly susceptible to contextual (pragmatic and task-related) factors, the default computation of negated words suggested by measures of neural activity is compatible with an incremental incorporation, as opposed to a two-stage processing that first affirms and then modifies to negate. As soon as it occurs, a negation marker changes (most likely reduces) the level of activity in brain areas representing the word in the scope of negation. Such an effect co-occurs with the recruitment of the neural mechanisms for response inhibition. The link between negation and inhibition has just begun to be investigated and its dynamics remain unknown. This link might reflect a functional coupling between two systems, which takes place in only certain, limited contexts and uses of negations, leaving aside the multiple and complex processes associated with logical negation (e.g. the inferences and virtually infinite set of representations that might follow an utterance as simple as ‘I don’t eat’ or ‘not many’ or ‘not short’ and so on).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

      



Future research should also ask whether this coupling generalizes to other semantic categories, beside the concrete action-related language. The motor system is closely related to the system for response inhibition, which could make the link particularly strong or even unique to the case of action-related language (see Beltrán et al. ). Moreover, it has repeatedly been noted that motor areas might not be a central component of the network that supports language understanding. This concern is raised by research on individuals with damaged motor brain centers (Papeo and Hochmann ) or impoverished motor abilities and experience (Vannuscorps and Caramazza ), who prove perfectly able to understand actions and action-related language. This circumstance makes particularly urgent the investigation of negation in other aspects of the brain network for language understanding, beyond the motor areas. How are the more general and abstract aspects of meanings, represented in regions such as the lateral temporal cortex (Nastase and Haxby ; Papeo et al. ; Wurm and Caramazza ), affected by negation? Another fundamental issue concerns the involvement of brain networks outside the ones implicated in language processing. The effects within the lexical-semantic network may have counterpart-effects in brain networks that support logical inference and reasoning in general (Reverberi et al. ; Camille et al. ). Are the neural mechanisms underlying sentential/lexical negation specific to language? Latest research suggests that precursors of negation (i.e. reasoning abilities required to process negation) may be independent from language (Cesana-Arlotti et al. ; Hochmann, Mody, and Carey ; Hochmann in this volume). While contributing to define new questions for future research, the body of studies reviewed here has set up a landmark with the characterization of the word not as a signal to the brain that changes the course of information processing and begins the transformation of a meaning (writing) in another meaning (not writing).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

R

.....................................

Aarons, Debra (). Aspects of the Syntax of American Sign Language. Ph.D. dissertation, Boston University, Boston. Abbott, Barbara (). Where have some of the presuppositions gone? In Drawing the Boundaries of Meaning: Neo-Gricean Studies in Pragmatics and Semantics in Honor of Laurence R. Horn : –. Abbott, Clifford (). Oneida. Munich: Lincom Europa. Abels, Klaus (). Who gives a damn about minimizers in questions? In Robert B. Young and Yuping Zhou (eds.), Proceedings of Semantics and Linguistic Theory : –. Ithaca, NY: Cornell University. Abels, Klaus, and Luisa Martí (). A unified approach to split scope. Natural Language Semantics  (): –. Abner, Natasha, and Ronnie Wilbur (). Quantification in American Sign Language. In Denis Paperno and Edward L. Keenan (eds.), Handbook of Quantifiers in Natural Language, Vol. , Studies in Linguistics and Philosophy , –. Berlin: Springer. Abraham, Werner (). The synchronic and diachronic semantics of German temporal noch and schon, with aspects of English still, yet, and already. Studies in Language : –. Abrusán, Márta (). Presuppositional and negative islands: A semantic account. Natural Language Semantics : –. Abrusán, Márta (). Weak Island Semantics. Oxford: Oxford University Press. Abrusán, Márta (). Presupposition cancellation: Explaining the “soft–hard” trigger distinction. Natural Language Semantics  (): –. Abrusán, Márta, and Benjamin Spector (). A semantics for degree questions based on intervals: Negative islands and their obviation. Journal of Semantics : –. Abusch, Dorit (). Lexical alternatives as a source of pragmatic presuppositions. In B. Jackson (ed.), Proceedings of SALT . Ithaca, NY: CLC Publications. Abusch, Dorit (). Presupposition triggering from alternatives. Journal of Semantics  (): –. Académie Tahitienne (). Grammaire de la langue tahitienne. Pape’ete: Fare Vana’a. Acquaviva, Paolo (). Operator composition and the licensing of negative concord. Geneva Generative Papers  (): –. Acquaviva, Paolo (). The Logical Form of Negation: A Study of Operator-Variable Structures in Syntax. New York: Garland. Adger, David, and George Tsoulas (). Circumstantial adverbials and aspect. In Jennifer R. Austin, Stefan Engelberg, and Gisa Rauh (eds.), The Syntax of Adverbials, –. Amsterdam: John Benjamins. Adolph, K. E., K. S. Kretch, and V. LoBue (). Fear of heights in infants? Current Directions in Psychological Science  (): –. Adriaanse, M. A., J. M. van Oosten, D. T. de Ridder, J. B. de Wit, and C. Evers (). Planning what not to eat: Ironic effects of implementation intentions negating unhealthy habits. Personality and Social Psychology Bulletin : –. Aguiar, Andréa, and Renée Baillargeon (). .-month-old infants’ reasoning about when objects should and should not be occluded. Cognitive Psychology : –. Akinlabi, Akinbiyi (). Featural affixation. Journal of Linguistics : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Alemanno, F., E. Houdayer, M. Cursi, S. Velikova, M. Tettamanti, G. Comi, S. F. Cappa, and L. Leocani (). Action-related semantic content and negation polarity modulate motor areas during sentence reading: An event-related desynchronization study. Brain Research : –. Alexander, Ronelle (). Bosnian, Croatian, Serbian: A Grammar with Sociolinguistic Commentary. Madison, WI: The University of Wisconsin Press. Alonso-Ovalle, Luis, and Paula Menéndez-Benito (). Epistemic indefinites: An overview. In Luis Alonso-Ovalle and Paula Menéndez-Benito (eds.), Epistemic Indefinites: Exploring Modality beyond the Verbal Domain, –. Oxford: Oxford University Press. Alrenga, Peter, and Christopher Kennedy (). No more shall we part: Quantifiers in English comparatives. Natural Language Semantics : –. Altmann, Gerry (). Language can mediate eye movement control within  milliseconds, regardless of whether there is anything to move the eyes to. Acta Psychologica : –. Alxatib, Sam and Francis Jeffry Pelletier (). The psychology of vagueness: Borderline cases and contradictions. Mind & Language : –. Amunts, Katrin, and Karl Zilles (). Architectonic mapping of the human brain beyond Brodmann. Neuron : –. Amunts, Katrin, Axel Schleicher, Uli Bürgel, Hartmut Mohlberg, Harry B. Uylings, and Karl Zilles (). Broca’s region revisited: Cytoarchitecture and intersubject variability. Journal of Comparative Neurology  (): –. Amunts, Katrin, Peter H. Weiss, Hartmut Mohlberg, Peter Pieperhoff, Simon Eickhoff, Jennifer M. Gurd, John C. Marshall, Nadim J. Shah, Gereon R. Fink, and Karl Zilles (). Analysis of neural mechanisms underlying verbal fluency in cytoarchitectonically defined stereotaxic space— the roles of Brodmann areas  and . Neuroimage  (), –. Amunts, Katrin, Olga Kedo, M. Kindler, Peter Pieperhoff, Hartmut Mohlberg, Nadim J. Shah, Ute Habel, Frank Schneider, and Karl Zilles (). Cytoarchitectonic mapping of the human amygdala, hippocampal region and entorhinal cortex: Intersubject variability and probability maps. Anatomy and Embryology : –. Anderbois, Scott (). Issues and Alternatives. Dissertation, University of California Santa Cruz. Andersen, Henning (). Abductive and deductive change. Language : –. Anderson, Catherine (). The structure and real-time comprehension of quantifier scope ambiguity. Ph.D. dissertation, Northwestern University. Anderson, M. C., and C. Green (). Suppressing unwanted memories by executive control. Nature  (): . Anderson, S., S. Huette, T. Matlock, and M. J. Spivey (). On the temporal dynamics of negated perceptual simulations. In F. Parrill, V. Tobin, and M. Turner (eds.), Mind, Form, and Body, –. Stanford, CA: CSLI Publications. Anscombe, G. E. M. (). Aristotle and the sea battle. Mind : –. Anscombe, G. E. M. (). Before and after. The Philosophical Review  (): –. Anselm, Saint (). The Lambeth fragments. In F. S. Schmitt (ed.), Ein neues unvollendetes Werk des hl. Anselm von Canterbury. Münster i. W.: Aschendorff. Anttila, Arto (). Deriving variation from grammar: A study of Finnish genitives. In F. Hinskens, R. van Hout, and W. L. Wetzels (eds.), Variation, Change, and Phonological Theory, –. Amsterdam: John Benjamins. Antzakas, Klimis (). Use of negative head movements in Greek Sign Language. In Ulrike Zeshan (ed.), Interrogative and Negative Constructions in Sign Languages, –. Nijmegen: Ishara Press. Aravena, P., Y. Delevoye-Turrell, V. Deprez, A. Cheylus, Y. Paulignan, V. Frak, and T. Nazir (). Grip force reveals the context sensitivity of language-induced motor activity during “action words” processing: Evidence from sentential negation. PLoS One . e. Arensen, Jonathan E. (). Murle Grammar. Juba, Sudan: SIL and University of Juba. Aristotle (). Categories and De Interpretatione, ed. and trans. J. L. Ackrill. Oxford: Clarendon. Aristotle (). Metaphysics, ed. H. Apostle. Bloomington: Indiana University Press.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Aristotle (). Prior Analytics, trans. Hugh Tredinnick. Cambridge, MA: Harvard University Press. Aristotle () Sophistici Elenchi, translated as On Sophistical Refutations by W. A. PickardCambridge. Online book. . Armstrong, A., and Z. Dienes (). Subliminal understanding of negation: Unconscious control by subliminal processing of word pairs. Consciousness and Cognition : –. Armstrong, Meghan E. (). Accounting for intonational form and function in Puerto Rican Spanish polar questions. Probus : –. Armstrong, Meghan E., and Scott Schwenter (). Prosody, accessibility, and sentential negation in Brazilian Portuguese. Annual Meeting of the Berkeley Linguistics Society  (): –. Aron, A. R., T. W. Robbins, and R. A. Poldrack (). Inhibition and the right inferior frontal cortex: One decade on. Trends Cogn. Sci.  (): –. Aronoff, Mark, Irit Meir, and Wendy Sandler (). The paradox of sign language morphology. Language : –. Arrotéia, Jéssica (). O Papel da Marcação Não Manual nas Sentenças Negativas em Língua de Sinais Brasileira (LSB). MA Thesis, Universidade Estadual de Campinas. Arroyo, F. V. (). Negatives in context. Journal of Verbal Learning and Verbal Behavior : –. Asher, Nicholas, and Brian Reese (). Negative bias in polar questions. In E. Maier, C. Bary, and J. Huitink (eds.), Proceedings of SuB: –. Asher, Nicholas, and Brian Reese (). Intonation and discourse: Biased questions. In S. Ishihara, S. Jannedy, and A. Schwarz (eds.), Interdisciplinary Studies on Information Structure , –. Potsdam: University of Potsdam. Atkinson, Joan, Ruth Campbell, Jane Marshall, Alice Thacker, and Bencie Woll (). Understanding ‘not’: Neuropsychological dissociations between hand and head markers of negation in BSL. Neuropsychologia : –. Atlas, Jay D. (). Negation, ambiguity, and presupposition. Linguistics and Philosophy  (): –. Atlas, Jay D. (). Negative quantifier noun phrases: A typology and an acquisition hypothesis. In Jack Hoeksema, Hotze Rullmann, Víctor Sánchez-Valencia, and Ton van der Wouden (eds.), Perspectives on Negation and Polarity Items, –. Amsterdam: John Benjamins. Atlas, Jay D. (). Presupposition. In L. R. Horn and G. Ward (eds.), The Handbook of Pragmatics, –. Blackwell. Attal, Pierre (). Négation de phrase et négation de constituant. Langue française : –. Auersperg, Alice M. I., Birgit Szabo, Auguste M. P. von Bayern, and Thomas Bugnyar (). Object permanence in the Goffin cockatoo (Cacatua goffini). Journal of Comparative Psychology : –. Auger, Julie, and Anne-José Villeneuve (). Using comparative sociolinguistics to inform European minority language policies: Evidence from contemporary Picard and regional French. Canadian Journal of Linguistics  (): –. Austin, Jennifer, Maria Blume, and Liliana Sánchez (). Morphosyntactic attrition in the L of Spanish–English bilingual children. Hispania : –. Austin, K., A. Theakston, E. Lieven, and M. Tomasello (). Young children’s understanding of denial. Developmental Psychology  (): . Autry, K. S., and W. H. Levine (). Activation of negated and non-negated entities. Journal of Pragmatics : –. Autry, K. S., and W. H. Levine (). Presupposition processing and the (re) activation of negated concepts. Discourse Processes : –. Bahlmann, Jörg, Jutta L. Mueller, Michiru Makuuchi, and Angela D. Friederici (). Perisylvian functional connectivity during processing of sentential negation. Frontiers in Psychology, Article . Baillargeon, R., E. S. Spelke, and S. Wasserman (). Object permanence in five-month-old infants. Cognition  (): –. Baka, J. (). Questionnaire kisuundi. Manuscript. Baker, C. L. (). Double negatives. Linguistic Inquiry : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Baker-Shenk, Charlotte (). A Micro-Analysis of the Nonmanual Components of Questions in American Sign Language. Ph.D. dissertation, University of California, Berkeley. Bald, Wolf-Dietrich (). The scope of negation and copula sentences in English. Journal of English Linguistics : –. Bar-Asher Siegal, Elitzur (a). Arguments in favor of an ambiguist approach to negation. Paper presented at the Göttingen workshop on negation and polarity. Bar-Asher Siegal, Elitzur (b). The case for external sentential negation: Evidence from Jewish Babylonian Aramaic. Linguistics  (): –. Bar-Asher Siegal, Elitzur, and Karen De Clercq (). From negative cleft to external negator: Eastern Aramaic lāw and Sicilian (Mussomeli) neca. Paper presented at DiGS . Bar-Asher Siegal, Elitzur, and Karen De Clercq (). From negative cleft to external negator. In A. Breitbarth et al. (eds.), Cycles in Language Change, –. Oxford: Oxford University Press. Barker, Chris (). Quantificational binding does not require c-command. Linguistic Inquiry : –. Barnes, Janet (). Tuyuca. In Kahrel and van den Berg (eds.), Typological Studies in Negation, –. Barres, P., and P. N. Johnson-Laird (). On imagining what is true (and what is false). Thinking and Reasoning : –. Barsalou, Lawrence, W. (). Perceptual symbol systems. Behavioral and Brain Sciences : –. Bartels, Christine (). The Intonation of English Statements and Questions. New York: Garland. Bartoli, E., A. Tettamanti, P. Farronato, A. Caporizzo, A. Moro, R. Gatti, D. Perani, and M. Tettamanti (). The disembodiment effect of negation: Negating action-related sentences attenuates their interference on congruent upper limb movements. Journal of Neurophysiology : –. Bartsch, Renate (). “Negative transportation” gibt es nicht. Linguistische Berichte : –. Bartsch, Renate, and Theo Vennemann (). Semantic Structures. Frankfurt a.M.: Athenäum. Barwise, Jon (). Scenes and other situations. The Journal of Philosophy, : –. Barwise, Jon, and Robin Cooper (). Generalized quantifiers and natural language. Linguistics and Philosophy : –. Barwise, Jon, and John Perry (). Semantic innocence and uncompromising situations. In P. A. French, T. Uehling, and H. K. Wettstein (eds.), Foundations of Analytic Philosophy (Midwest Studies in Philosophy, Vol. ), –. Minneapolis: University of Minnesota Press. Barwise, Jon, and John Perry (/). Situations and Attitudes. Stanford, CA: CSLI Publications. Bastiaanse, Roelien, Gosse Bouma, and Wendy Post (). Linguistic complexity and frequency in agrammatic speech production. Brain and Language  (): –. Bates, Douglas, Martin Mächler, Ben Bolker, and Steve Walker (). Fitting linear mixed-effects models using lme. Journal of Statistical Software : –. Batllori, Montserrat (). The significance of formal features in language change theory and the evolution of minimizers. In P. Larrivée and C. Lee (eds.), Negation and Polarity: Experimental Perspectives, –. Cham: Springer. Batllori, Montserrat, and M. Lluïsa Hernanz (). Emphatic polarity particles in Spanish and Catalan. Lingua : –. Bauer, Winifred (). Maori ko again. Te Reo : –. Bauer, Winifred (with William Parker and Te Kareongawai Evans) (). Maori. London and New York: Routledge. Baunaz, Lena, and Eric Lander (). Nanosyntax: The basics. In Lena Baunaz, Karen De Clercq, Liliane Haegeman, and Eric Lander (eds.), Exploring Nanosyntax, –. Oxford: Oxford University Press. Bayer, Josef (). What Bavarian negative concord reveals about the syntactic structure of German. In Juan Mascaró and Marina Nespor (eds.), Grammar in Progress, –. Dordrecht: Foris Publications.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Bayer, Josef (). Nominal negative quantifiers as adjuncts. The Journal of Comparative Germanic Linguistics  (): –. Beaupoil-Hourdel, Pauline, Dominique Boutet, and Aliyah Morgenstern (). A child’s multimodal negations from  to : The interplay between modalities. In Pierre Larrivée and Chungmin Lee (eds.), Negation and Polarity: Experimental Perspectives, Language, Cognition, and Mind, –. Zurich: Springer International Publishing Switzerland. Beaver, David (). The kinematics of presupposition. In P. Dekker and M. Stokhof (eds.), Proceedings of the Eighth Amsterdam Colloquium, –. Institute of Logic, Language and Computation Publications, Amsterdam. Beaver, David (). Presupposition. In J. van Benthem and A. ter Meulen (eds.), Handbook of Logic and Language, –. Amsterdam: North-Holland. Beaver, David (). Presupposition and Assertion in Dynamic Semantics, Vol. . Stanford, CA: CSLI publications. Beaver, David, and Brady Clark (). Always and only: why not all focus-sensitive operators are alike. Natural Language Semantics : –. Beaver, David, and Brady Clark (). Sense and Sensitivity: How Focus Determines Meaning. Cambridge: Wiley-Blackwell. Beaver, David and Emiel Krahmer (). A partial account of presupposition projection. Journal of Logic, Language and Information  (): . Beck, Sigrid (). Quantified structures as barriers for LF movement. Natural Language Semantics : –. Beck, Sigrid (). Intervention effects follow from focus interpretation. Natural Language Semantics : –. Beck, Sigrid, and Shin-Sook Kim (). On wh- and operator scope in Korean. Journal of East Asian Linguistics : –. Beck, Sigrid, and Shin-Sook Kim (). Intervention effects in alternative questions. Journal of Comparative Germanic Linguistics : –. Beck, Sigrid, and Hotze Rullmann (). A flexible approach to exhaustivity in questions. Natural Language Semantics : –. Beghelli, Filippo (). The Phrase Structure of Quantifier Scope: UCLA dissertation. Beghelli, Filippo, and Tim Stowell (). Distributivity and negation: The syntax of each and every. In A. Szabolcsi (ed.), Ways of Scope Taking, –. Dordrecht: Springer. Beheydt, G. (). Het gebruik en de vorm van de negatie in het zuidelijke Nederlands in een diachronisch perspectief (e–e eeuw). Licentiate dissertation, University Leuven. Belletti, Adriana (). Generalized Verb Movement: Aspects of Verb Syntax. Turin: Rosenberg and Sellier. Belletti, Adriana (). Verb positions; NP positions: Evidence from Italian. In David Lightfoot and Norbert Hornstein (eds.), Verb Movement, –. Cambridge: Cambridge University Press. Belletti, Adriana (). Inversion as focalization. In Aafke Hulk and Jean-Yves Pollock (eds.), Subject Inversion in Romance and the Theory of Universal Grammar, –. New York: Oxford University Press. Belletti, Adriana (). Aspects of the Low IP Area. In Luigi Rizzi (ed.), The Structure of IP and CP: The Cartography of Syntactic Structures, Vol. , – . Oxford: Oxford University Press. Bellugi, Ursula (). The Acquisition of the System of Negation in Children’s Speech. Doctoral dissertation, Harvard University. Beltrán, David, Mercedes Muñetón-Ayala, and Manuel de Vega (). Sentential negation modulates inhibition in a stop-signal task: Evidence from behavioral and ERP data. Neuropsychologia : –, . Beltrán, David, Yurena Morera, Enrique García-Marco, and Manuel de Vega (). Brain inhibitory mechanisms are involved in the processing of sentential negation, regardless of its content: Evidence from EEG theta and beta rhythms. Frontiers in Psychology, : .

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Bendall, Kent (). Negation as a sign of negative judgment. Notre Dame Journal of Formal Logic  (): –. Benincà, P., and C. Poletto (). Topic, Focus, and V: Defining the CP sublayers. In L. Rizzi (ed.), The Structure of CP and IP: The Cartography of Syntactic Structures, –. Oxford and New York: Oxford University Press. Benitez-Quiroz, Fabian C., Ronnie B. Wilburb, and Aleix M. Martinez (). The not face: A grammaticalization of facial expressions of emotion. Cognition : –. Ben-Shachar, Michal, Dafna Palti, and Yosef Grodzinsky (). Neural correlates of syntactic movement: Converging evidence from two fMRI experiments. NeuroImage  (): –. Bentzen, Kristine (). Embedded root phenomena, assertion, presupposition, and main point of utterance. Paper presented at the ZAS Workshop on Root Phenomena, Berlin, September –, . . Beran, Michael J., and J. David Smith (). Information seeking by rhesus monkeys (Macaca mulatta) and capuchin monkeys (Cebus apella). Cognition : –. Bermúdez, José L. (). Thinking without Words. Oxford: Oxford University Press. Bermúdez, José L. (). Animal reasoning and proto-logic. In Susan L. Hurley and Matthew Nudds (eds.), Rational Animals?, –. Oxford: Oxford University Press. Bernardi, Raffaella. (). Reasoning with Polarity in Categorial Type Logic. Ph.D. thesis, University of Utrecht. Bernini, Giuliano, and Paolo Ramat (). La frase negativa nelle lingue d’Europa. Bologna: Il Mulino. Bernini, Giuliano, and Paolo Ramat (). Negative Sentences in the Languages of Europe: A Typological Approach. Berlin: Mouton de Gruyter. Besson, Corine (). Logical knowledge and ordinary reasoning. Philosophical Studies  (): –. Besten, Hans den (). Double negation and the genesis of Afrikaans. In Pieter Muyskens and Norval Smith (eds.), Substrata versus Universals in Creole Languages, –. Amsterdam: John Benjamins. Beyer, Klaus (). Double negation-marking: A case of contact-induced grammaticalization in West Africa? In Cyffer et al. (eds.), Negation Patterns in West African Languages and Beyond, –. Amsterdam: Benjamins. Béziau, Jean-Yves, and Basti Gianfranco (eds.) (). The Square of Opposition: A Cornerstone of Thought. Basel: Birkhäuser. Biberauer, Theresa (). Competing reinforcements: When languages opt out of Jespersen’s Cycle. In A. van Kemenade and N. de Haas (eds.), Historical Linguistics : Selected Papers from the th International Conference on Historical Linguistics, Nijmegen, – August , –. Amsterdam: John Benjamins. Biberauer, Theresa (). Jespersen off course? The case of contemporary Afrikaans negation. In Elly van Gelderen (ed.), Cyclical Change (Linguistik Aktuell/Linguistics Today ), –. Amsterdam: John Benjamins. Biberauer, Theresa (). The Final-over-Final Condition and particles. In Michelle Sheehan, Theresa Biberauer, Anders Holmberg, and Ian Roberts (eds.), The Final-over-Final Condition: A Syntactic Universal, –. Cambridge, MA: MIT Press. Biberauer, Theresa, and Sonia Cyrino (). Negative developments in Afrikaans and Brazilian Portuguese. Paper presented at the th Colloquium on Generative Grammar (CGG) conference (Vitoria). Biberauer, Theresa, and Ian Roberts (). Comments on Jäger’s “Anything is nothing is something”: On the diachrony of polarity types of indefinites. Natural Language and Linguistic Theory  (): –. Biberauer, Theresa, and Ian Roberts (). Negative words and related expressions: A new perspective on some familiar puzzles. In P. Larrivée and R. Ingham (eds.), The Evolution of Negation: Beyond the Jespersen Cycle, –. Berlin: de Gruyter.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Biberauer, Theresa, and Hedde Zeijlstra (). Negative concord in Afrikaans: Filling the typological gap. Journal of Semantics : –. Biberauer, Theresa, Anders Holmberg, and Ian Roberts (), A syntactic universal and its consequences. Linguistic Inquiry  (): –. Biezma, Maria, and Kyle Rawlins (). Rhetorical questions: Severing questioning from asking. In Dan Burgdorf, Jacob Collard, Sireemas Maspong, and Brynhildur Stefánsdóttir (eds.), Proceedings of Semantics and Linguistic Theory , –. Washington, DC: LSA Open Journal Systems. Binder, J. R., and R. H. Desai (). The neurobiology of semantic memory. Trends in Cognitive Sciences : –. Binder, J. R., R. H. Desai, W. W. Graves, and L. L. Conant (). Where is the semantic system? A critical review and meta-analysis of  functional neuroimaging studies. Cerebral Cortex  (): –. Birner, Betty J. (). Semantic and pragmatic contributions to information status. Acta Lingvistica Hafniensia : –. Blackings, Mairi, and Nigel Fabb (). A Grammar of Ma’di. Berlin: Mouton de Gruyter. Blaisdell, Aaron P., Kosuke Sawa, Kenneth J. Leising, and Michael R. Waldmann (). Causal reasoning in rats. Science : –. Blaisdell, Aaron P., Kenneth J. Leising, W. David Stahlman, and Michael R. Waldmann (). Rats distinguish between absence of events and lack of information in sensory preconditioning. International Journal of Comparative Psychology : –. Blakemore, Diane (). Semantic Constraints on Relevance. Oxford: Blackwell. Blanchette, Frances (). Negative concord in English. Linguistic Variation : –. Blanchette, Frances (). English negative concord, negative polarity, and double negation. Doctoral Dissertation, CUNY Graduate Center. Blanchette, Frances (). Micro-syntactic variation in American English Negative Concord. Glossa: A Journal of General Linguistics  (): –. Blanchette, Frances, and Cynthia Lukyanenko (). Unacceptable grammars? An eye-tracking study of English Negative Concord. Language and Cognition  (): –. Blanchette, Frances, and Marianna Nadeu (). Prosody and the meanings of English negative indefinites. Journal of Pragmatics : –. Blanchette, Frances, Marianna Nadeu, Jeremy Yeaton, and Viviane Déprez (). English negative concord and double negation: The division of labor between syntax and pragmatics. In Proceedings of the Annual Meeting of the Linguistic Society of America : –. Blaszczak, Joanna (). Towards a binding analysis of negative polarity items. Linguistics in Potsdam : –. [Also appeared in: Kosta, P. and J. Frasek (eds.) (), Current Approaches to Formal Slavic Linguistics, –. Frankfurt am Main: Peter Lang.] Blaxter, Tam, and David Willis (). Pragmatic Differentiation of Negative Markers in the Early Stages of Jespersen’s Cycle in North Germanic. Amsterdam: Diachronica. Bley-Vroman, Robert (). What is the logical problem of foreign language learning? In S. Gass and J. Schachter (eds.), Linguistic Perspectives on Second Language Acquisition, –. Cambridge: Cambridge University Press. Blochowiak, Joanna, and Cristina Grisot (). The pragmatics of descriptive and metalinguistic negation: Experimental data from French. Glossa: A Journal of General Linguistics  (): –. . Bloom, Lois (). Language Development: Form and Function in Emerging Grammars. Cambridge, MA: MIT Press. Blutner, Reinhard (). Lexical pragmatics. Journal of Semantics : –. Blutner, Reinhard (). Some aspects of optimality in natural language interpretation. Journal of Semantics : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Bochvar, D. (). Ob odnom tréhznačnom isčislénii i égo priméneii k analizu paradoksov klassičéskogo rasširénnogo funkcional‘nogo isčislénijá. Matématičéskij sbornik : –. (English translation (): On a three-valued logical calculus and its applications to the analysis of the paradoxes of the classical extended functional calculus. History and Philosophy of Logic : –.) Boeckx, Cedric (). Scope reconstruction and A-movement. Natural Language and Linguistic Theory : –. Boër, S. E., and W. G. Lycan (). The Myth of Semantic Presupposition. Bloomington, IN: Indiana Linguistics Club. Bohn, Manuel, Josep Call, and Michael Tomasello (). Communication about absent entities in great apes and human infants. Cognition : –. Bohn, Manuel, Josep Call, and Michael Tomasello (). The role of past interactions in great apes’ communication about absent entities. Journal of Comparative Psychology : –. Bohn, Manuel, Matthias Allritz, Josep Call, and Christoph Völter (). Information seeking about tool properties in great apes. Scientific Reports : . Bokura, Hirokazu, Shuhei Yamaguchi, and Shotai Kobayashi (). Electrophysiological correlates for response inhibition in a Go/NoGo task. Clinical Neurophysiology : –. Bolinger, Dwight (). Contrastive accent and contrastive stress. Language : –. Bolinger, Dwight (). Forms of English: Accent, Morpheme, Order. Cambridge, MA: Harvard University Press. Bolinger, Dwight (). Postposed main phrases: An English rule for the Romance subjunctive. Canadian Journal of Linguistics : –. Bolinger, Dwight (). Degree Words. The Hague: Mouton. Bolinger, Dwight (). Meaning and Form. London: Longman. Bolinger, Dwight (). Yes-no questions are not alternative questions. In Henry Hiz (ed.), Questions, –. Dordrecht: Reidel. Bond, Oliver (). Negation through reduplication and tone: Implications for the Lexical Functional Grammar/Paradigm Function Morphology interface. Journal of Linguistics  (): –. Boone, Enrico (a). Exceptional movement is unexceptionally local. In Jyoti Iyer and Leland Kusmer (eds.), Proceedings of NELS : –. Amherst, MA: GLSA. Boone, Enrico (b). The Syntax and Licensing of Gapping and Fragments. Universiteit Leiden dissertation. Bor, Daniel, and Adrian M. Owen (). A common prefrontal-parietal network for mnemonic and mathematical recoding strategies within working memory. Cerebral Cortex  (): –. Bor, Daniel, Nick Cumming, Catherine E. L. Scott, and Adrian M. Owen (). Prefrontal cortical involvement in verbal encoding strategies: Modality independent PFC involvement in strategies. European Journal of Neuroscience  (): –. Borer, Hagit (). The Normal Course of Events. Oxford: Oxford University Press. Borkin, Ann (). Polarity items in questions. In Douglas Adams, Mary Ann Campbell, Victor Cohen, Julie Lovins, Edward Maxwell, Carolyn Nygren, and John Reighard (eds.), Papers from the th Regional Meeting of the Chicago Linguistic Society, –. Chicago: CLS. Bornkessel, Ina, Stefan Zysset, Angela D. Friederici, D. Yves von Cramon, and Matthias Schlesewsky (). Who did what to whom? The neural basis of argument hierarchies during language comprehension. NeuroImage  (): –. Borsley, Robert D., and Bob Morris Jones (). Welsh Negation and Grammatical Theory. Cardiff: University of Wales Press. Bosanquet, Bernard (). Logic, Vol. . Oxford: Clarendon. Bošković, Željko, and Jon Gajewski (). Semantic correlates of the NP/DP parameter. NELS : –. Botvinick Todd, M. M. Braver, D. M. Barch, C. S. Carter, J. D. Cohen, M. M. Botvinick, and T. S. Braver (). Conflict monitoring and cognitive control. Psychological Review : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Boves, Lou, and Nelleke Oostdijk (). Spontaneous speech in the Spoken Dutch Corpus. In Proceedings ISCA & IEEE Workshop on Spontaneous Speech Processing and Recognition (SSPR), –. Japan: Tokyo. Brandom, Robert (). Making It Explicit. Cambridge, MA: Harvard University Press. Brasoveanu, Adrian, Donka Farkas, and Floris Roelofsen (). N-words and sentential negation: Evidence from polarity particles and VP ellipsis. Semantics & Pragmatics  (): –. Brasoveanu, Adrian, Karen De Clercq, Donka Farkas, and Floris Roelofsen (). Question tags and sentential negativity. Lingua : –. Bräuer, Juliane, Josep Call, and Michael Tomasello (). Visual perspective taking in dogs (Canis familiaris) in the presence of barriers. Applied Animal Behaviour Science : –. Bray, Joel, Christopher Krupenye, and Brian Hare (). Ring-tailed lemurs (Lemur catta) exploit information about what others can see but not what they can hear. Animal Cognition : –. Breheny, Richard (). Exhaustivity, homogeneity, and definiteness. In Paul Dekker and Michael Franke (eds.), Fifteenth Amsterdam Colloquium, –. Amsterdam: ILLC. Breitbarth, Anne (). A hybrid approach to Jespersen’s Cycle in West Germanic. Journal of Comparative Germanic Linguistics : –. Breitbarth, Anne (a). Negation in the history of Low German and Dutch. In David Willis, Christopher Lucas, and Anne Breitbarth (eds.), The History of Negation in the Languages of Europe and the Mediterranean. Vol. : Case Studies, –. Oxford: Oxford University Press. Breitbarth, Anne (b). Indefinites, negation and Jespersen’s Cycle in the history of Low German. Diachronica : –. Breitbarth, Anne (). The History of Low German Negation. Oxford: Oxford University Press. Breitbarth, Anne (). Jespersen’s cycle = minimize structure + feature economy. In Silvio Cruschina, Katharina Hartmann, and Eva-Maria Remberger (eds.), Studies on Negation. Syntax, Semantics, and Variation, –. Göttingen: V+R unipress/Vienna University Press. Breitbarth, Anne, and Liliane Haegeman (). Continuity is change: The long tail of Jespersen’s cycle in Flemish. In Anne Breitbarth et al. (eds.), Continuity and Change in Grammar, –. Dordrecht: Benjamins. Breitbarth, Anne, and Liliane Haegeman (). The distribution of preverbal en in (West) Flemish: Syntactic and interpretive properties. Lingua : –. Breitbarth, Anne, Christopher Lucas, and David Willis (). Incipient Jespersen’s Cycle: The (non?) grammaticalization of new negative markers. In Jürg Fleischer and Horst Simon (eds.), Comparing Diachronies/Sprachwandelvergleich, –. Tübingen: Niemeyer. Britain, David (). Diffusion, leveling, simplification and reallocation in past tense BE in the English Fens. Journal of Sociolinguistics : –. Brodmann, Korbinian (). Vergleichende Lokalisationslehre der Grosshirnrinde in ihren Prinzipien dargestellt auf Grund des Zellenbaues. Barth. Bromberger, Sylvain (). On What We Know We Don’t Know: Explanation, Theory, Linguistics, and how Questions Shape them. Chicago: Chicago University Press. Broschart, Jürgen (). Negation in Tongan. In Hovdhaugen and Mosel (eds.), Negation in Oceanic Languages, –. Brown, C., and P. Hagoort (). The processing of the N: Evidence from masked priming. Journal of Cognitive Neuroscience : –. Brown, D. Richard (). Kresh. In Kahrel and van den Berg (eds.), Typological Studies in Negation, –. Brown, L., and P. Prieto (in press). Gesture and prosody as sister systems in multimodal communication. In M. Haugh, D. Kadar, and M. Terkourafi (eds.), Handbook of Sociopragmatics. Brown, Penelope, and Stephen C. Levinson (). Politeness: Some Universals in Language Usage. Cambridge: Cambridge University Press. Brown, Roger (). A First Language: The Early Stages. Cambridge, MA: Harvard University Press.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Brown, Sue (). The Syntax of Negation in Russian: A Minimalist Approach. Stanford, CA: CSLI Publications. Brown, Sue, and Steven Franks (). Asymmetries in the scope of Russian Negation. Journal of Slavic Linguistics : –. Bruening, Benjamin (). QR obeys Superiority: ACD and frozen scope. Linguistic Inquiry  (): –. Bugnyar, Thomas, Stephan A. Reber, and Cameron Buckner (). Ravens attribute visual access to unseen competitors. Nature Communications : . Burge, Tyler (). Demonstrative constructions, reference and truth. Journal of Philosophy  (): –. Büring, Daniel (). The th Street Bridge Accent: On the Meaning of Topic and Focus. Ph.D. dissertation Universität Tübingen. Büring, Daniel (a). The Meaning of Topic and Focus: The th Street Bridge Accent. London and New York: Routledge. Büring, Daniel (b). The great scope inversion conspiracy. Linguistics and Philosophy : –. Büring, Daniel (). On D-trees, beans, and B-accents. Linguistics and Philosophy : –. Büring, Daniel (). Negative inversion. In Leah Bateman and Cherlon Ussery (eds.), Proceedings of the th Annual Meeting of the North East Linguistic Society, –. Büring, Daniel (). More or less. In Proceedings from the Annual Meeting of the Chicago Linguistic Society  (): –. Büring, Daniel (). Intonation and Meaning. Oxford: Oxford University Press. Büring, Daniel, and Christine Gunlogson (). Aren’t positive and negative polarity questions the same? Ms., UCSC. Burnett, Heather, and Anne Dagnac (). A tale of two negations in Picard: A dilemma for the formal analysis of negative concord. Paper presented at the SyHD Workshop, December , , Frankfurt, Germany. . Burnett, Heather, Hilda Koopman, and Sali Tagliamonte (). Structural explanations in syntactic variation: The evolution of English negative and polarity indefinites. Language Variation and Change  (): –. Burnett, Heather, Hilda Koopman, and Sali A. Tagliamonte (in press). Soft syntax and the evolution of negative and polarity indefinites in the history of English. Language Variation and Change. Burnett, Heather, Mireille Tremblay, and Helene Blondeau (). The variable grammar of negative concord in Montréal French. U. Penn Working Papers in Linguistics, Vol. .. Burnsky, J., E. Darley, H. Muller, J. Buffinton, and C. Phillips (). Interpreting Negation in Incomplete Propositions. Talk at th Annual CUNY Conference on Human Sentence Processing, Massachusetts Institute of Technology. Burton-Roberts, N. (a). The Limits to Debate: A Revised Theory of Semantic Presupposition. Cambridge: Cambridge University Press. Burton-Roberts, N. (b). On Horn’s dilemma: Presupposition and negation. Journal of Linguistics  (): –. Burton-Roberts, N. (). On preservation under negation. Lingua  (–): –. Burton-Roberts, N. (). Presupposition-cancellation and metalinguistic negation: A reply to Carston. Journal of Linguistics : –. Butler, Jonny (). A minimalist treatment of modality. Lingua : –. Cable, Seth (). The Grammar of Q: Q-Particles, Wh-Movement and Pied-Piping. Oxford and New York: Oxford University Press. Caha, Pavel (). The Nanosyntax of Case. Tromsø: University of Tromsø dissertation. Caha, Pavel and Guido Vanden Wyngaerd (eds.) (). Glossa Special Collection on *ABA. Glossa. Caha, Pavel, Karen De Clercq, and Guido Vanden Wyngaerd (). The fine structure of the comparative. Studia Linguistica: –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Call, Josep (). Inferences about the location of food in the great apes (Pan paniscus, Pan troglodytes, Gorilla gorilla, and Pongo pygmaeus), Journal of Comparative Psychology  (): –. Call, Josep (). Do apes know that they could be wrong? Animal Cognition  (): –. Call, Josep, and Malinda Carpenter (). Do apes and children know what they have seen? Animal Cognition : –. Call, Josep, and Michael Tomasello (). The Gestural Communication of Apes and Monkeys. New York: Lawrence Erlbaum Associates. Call, Josep, and Michael Tomasello (). Does the chimpanzee have a theory of mind?  years later. Trends in Cognitive Sciences : –. Cameron-Faulkner, Theo, Elena Lieven, and Anna Theakston (). What part of no do children not understand? A usage-based account of multiword negation. Journal of Child Language : –. Camille, Nathalie, Giorgio Coricelli, Jerome Sallet, Pascale Pradat-Diehl, Jean-René Duhamel, and Angela Sirigu (). The involvement of the orbitofrontal cortex in the experience of regret. Science : –. Cancino, Herlinda, Ellen Rosansky, and John Schumann (). The acquisition of English negatives and interrogatives by native Spanish speakers. In E. Hatch (ed.), Second Language Acquisition: A Book of Readings, –. Rowley, MA: Newbury House. Cantello, Roberto, Maria Gianelli, Carlo Civardi, and Roberto Mutani (). Magnetic brain stimulation: The silent period after the motor evoked potential. Neurology : –. Caponigro, Ivano, and Jon Sprouse (). Rhetorical questions as questions. In Estella PuigWaldmueller (ed.), Proceedings of Sinn und Bedeutung , –. Barcelona: UPF. Caramazza, A., and E. B. Zurif (). Dissociation of algorithmic and heuristic processes in language comprehension: Evidence from aphasia. Brain and Language  (): –. . Carden, Guy (). A note on conflicting idiolects. Linguistic Inquiry : –. Cardinaletti, Anna (). German and Italian modal particles and clause structure. The Linguistic Review : –. Cardinaletti, Anna, and Michal Starke (). The typology of structural deficiency: On three grammatical classes. University of Venice Working Papers in Linguistics  (): –. Reprint in Henk van Riemsdijk (ed.) (), Clitics in the Languages of Europe. Berlin: Mouton de Gruyter, –. Carey, Susan, (). The Origin of Concepts. Oxford: Oxford University Press. Carey, Susan, and E. Bartlett (). Acquiring a single new word. Proceedings of the Stanford Child Language Conference : –. Carlson, Greg N. (). Reference to Kinds in English. Ph.D. Dissertation. University of Massachusetts Amherst. Carlson, Greg N. (). Polarity any is existential. Linguistic Inquiry : –. Carlson, Greg N. (). Distribution of Free-Choice any. In Seventeenth Regional Meeting of the Chicago Linguistics Society, –. Carlson, Lauri (). Dialogue Games. Dordrecht: D. Reidel. Carpenter, Patricia A. (). Extracting information from counterfactual clauses. Journal of Verbal Learning and Verbal Behavior : –. Carpenter, Patricia A., and Marcel Adam Just (). Sentence comprehension: A psycholinguistic processing model of verification. Psychological Review  (): –. Carpenter, Patricia A., Marcel Adam Just, Timothy A. Keller, William F. Eddy, and Keith R. Thulborn (). Time course of fMRI-activation in language and spatial networks during sentence comprehension. Neuroimage  (): –. Carston, Robyn (). Metalinguistic negation and echoic use. Journal of Pragmatics  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Carston, Robyn (). Negation, “presupposition” and the semantics/pragmatics distinction. Journal of Linguistics  (): –. Carston, Robyn (). Negation, “presupposition” and metarepresentation: A response to Noël Burton-Roberts. Journal of Linguistics : –. Carston, Robyn (). Thoughts and Utterances: The Pragmatics of Explicit Communication. Oxford: Blackwell. Carston, Robyn, and Eun-Ju-Noh (), A truth-functional account of metalinguistic negation, with evidence from Korean. Language Sciences  (–): –. Cartoni, Bruno, and Marie-Aude Lefer (). Negation and lexical morphology across languages: Insights from a trilingual translation corpus. Poznań Studies in Contemporary Linguistics  (): –. Casati, Roberto, and Achille C. Varzi (). Events. In Edward Zalta (ed.), The Stanford Encylopedia of Philosophy, Online. Castañeda, Hector-Neri (). Comments. In Nicholas Rescher (ed.), The Logic of Decision and Action, –. Pittsburgh: University of Pittsburgh Press. Castroviejo Miró, Elena (). Deconstructing exclamations. Catalan Journal of Linguistics : –. Castroviejo Miró, Elena (). On question tags and confirmation requests. Talk at GLOW . Budapest, April , . Catalani, Luigi (). Die Negation im Mittelfranzösischen. Frankfurt am Main: Peter Lang. Cattell, Ray (). Negative transportation and tag questions. Language  (): –. Cechetto, Carlo (). Explaining the locality conditions of QR: Consequences for the theory of phases. Natural Language Semantics : –. Cepeda, Paola (). Negation and time: Against Expletive Negation in Temporal Clauses. Ph.D. dissertation. Stony Brook University. Cesana-Arlotti, Nicoló, Ana Martín, Ernő Téglás, Liza Vorobyova, Ryszard Cetnarski, and Luca L. Bonatti (). Precursors of logical reasoning in preverbal human infants. Science  (): –. Chaker, Salem, and Dominique Caubet (). La négation en berbère et en arabe maghrébin. Paris: L’Harmattan. Chambers C. D., H. Garavan, and M. A. Bellgrove (). Insights into the neural basis of response inhibition from cognitive and clinical neuroscience. Neurosci Biobehav Rev : –. Chapman, Siobhan (). Some observations on metalinguistic negation. Journal of Linguistics  (): –. Chase, William G., and Herbert H. Clark (). Semantics in the perception of verticality. British Journal of Psychology : –. Chase, William G., and Herbert H. Clark (). Mental operations in the comparison of sentences and pictures. In L. Gregg (ed.), Cognition in Learning and Memory. New York: Wiley. Chatzikonstantinou, Anastasios (). Semantic and Prosodic Processing of NPIs in Greek. Ph.D. dissertation, University of Chicago. Chatzopoulou, Katerina (). Negation and nonveridicality in the history of Greek. Ph.D. dissertation, University of Chicago. Chatzopoulou, Katerina (). The Greek Jespersen’s cycle: Renewal, stability and structural micro– elevation. In Chiara Gianollo, Agnes Jäger, and Doris Penka (eds.), Language Change at the Syntax–Semantics Interface, –. Berlin: de Gruyter Mouton. Chatzopoulou, Katerina (). Negation and Nonveridicality in the History of Greek. Oxford: Oxford University Press. Cheng, Lisa L.-S. and Yafei Li (). Double negation in Chinese and multi projections. Paper presented at the North American Conference on Chinese Linguistics, Cornell University. Cheshire, Jenny (). Variation in the use of ain’t in an Urban British English Dialect. Language in Society  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Chierchia, Gennaro (). Questions with quantifiers. Natural Language Semantics : –. Chierchia, Gennaro (). Scalar implicatures, polarity phenomena, and the syntax/pragmatics interface. In A. Belletti (ed.), Structures and Beyond: The Cartography of Syntactic Structures, –. Oxford: Oxford University Press. Chierchia, Gennaro (). Broaden your views: Implicatures of domain widening and the “logicality” of language. Linguistic Inquiry  (): –. Chierchia, Gennaro (). Dynamics of Meaning: Anaphora, Presupposition, and the Theory of Grammar. Chicago: University of Chicago Press. Chierchia, Gennaro (). Logic in Grammar: Polarity, Free Choice, and Intervention. Oxford Studies in Semantics and Pragmatics. Oxford: Oxford University Press. Chierchia, Gennaro, and Sally McConnell-Ginet (). Meaning and Grammar: An Introduction to Semantics. Cambridge, MA: MIT Press. Chierchia, Gennaro, Danny Fox, and Benjamin Spector (). Scalar implicature as a grammatical phenomenon. In C. Maienborn, K. von Heusinger, and P. Portner (eds.), Semantics: An International Handbook of Natural Language Meaning, Vol. , –. Berlin: Mouton de Gruyter. Childs, Claire (). Integrating syntactic theory and variationist analysis: The structure of negative indefinites in regional dialects of British English. Glossa: A Journal of General Linguistics  (): .–. Childs, Claire, Christopher Harvey, Karen Corrigan, and Sali Tagliamonte (forthcoming). Transatlantic perspectives on variation in negative expressions. English Language and Linguistics, –. Cho, Young-mee Yu and Ki-Sun Hong (). Evidence for the VP-constituent from child Korean. In Eve V. Clark and Yo Matsumoto (eds.), Papers and Reports on Child Language Development . Proceedings of the Annual Child Language Research Forum, –. Choi, Soonja (). The semantic development of negation: A cross-linguistic longitudinal study. Journal of Child Language  (): –. Chomsky, Noam (). Syntactic Structures. The Hague: Mouton. Chomsky, Noam (). Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Chomsky, Noam (). The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, Noam (). Minimalist inquiries: The framework. In Roger Martin et al. (eds.), Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik, –. Cambridge, MA: The MIT Press. Chomsky, Noam (). Derivation by phase. In Michael Kenstowicz (ed.), Ken Hale: A Life in Language, –. Cambridge, MA: The MIT Press. Chow, W., E. Lau, S. Wang, and C. Phillips (). Wait a second! Delayed impact of argument roles on online verb prediction. Language, Cognition, and Neuroscience : –. Christensen, Ken Ramshøj (). Interfaces: Negation—Syntax—Brain. Aarhus: Aarhus University Ph.D. dissertation. . Christensen, Ken Ramshøj (). Negative and affirmative sentences increase activation in different areas in the brain. Journal of Neurolinguistics  (): –. Christensen, Ken Ramshøj, and Mikkel Wallentin (). The locative alternation: Distinguishing linguistic processing cost from error signals in Broca’s region. NeuroImage  (): –. Christensen, Ken Ramshøj, Johannes Kizach, and Anne Mette Nyvad (). The processing of syntactic islands: An fMRI study. Journal of Neurolinguistics  (): –. Christianson, K., S. G. Luke, and F. Ferreira (). Effects of plausibility on structural priming. Journal of Experimental Psychology: Learning Memory and Cognition  (): –. Christianson, K., A. Hollingworth, J. F. Halliwell, and F. Ferreira (). Thematic roles assigned along the garden path linger. Cognitive Psychology  (): –. Chumbow, Beban S., and Pius N. Tamanji (). Bafut. In Kahrel and van den Berg (eds.), Typological Studies in Negation, –. Chung, Eun Sun (). Sources of difficulty in L scope judgments. Second Language Research  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Chung, Inkie (). Suppletive negation in Korean and distributed morphology. Lingua : –. Chung, Sandra (). On the gradual nature of syntactic change. In Charles N. Li (ed.), Mechanisms of Syntactic Change, –. Austin: University of Texas Press. Chung, Sandra (). Case Marking and Grammatical Relations in Polynesian. Austin: University of Texas Press. Chung, Sandra (). The Design of Agreement. Chicago: The University of Chicago Press. Chung, Sandra (). Reaffirming Māori negatives as verbs. Paper presented at the Workshop on the Syntax of Polynesian Languages, University of Toronto, June –. Churchward, C. Maxwell (). Tongan Grammar. Oxford: Oxford University Press. Chwilla, D. J., and H. H. J. Kolk (). Accessing world knowledge: Evidence from N and reaction time priming. Cognitive Brain Research  (): –. Ciardelli, Ivano, Jeroen Groenendijk, and Floris Roelofsen (). Inquisitive semantics: A new notion of meaning. Language and Linguistics Compass  (): –. Cinque, Guglielmo (). Mica. Annali della Facoltà di Lettere e Filosofia dell’Università di Padova : –. Reprinted in Cinque, Teoria linguistica e sintassi italiana, –. Bologna: Il Mulino, . Cinque, Guglielmo (). Types of Ā-Dependencies. Cambridge, MA: MIT Press. Cinque, Guglielmo (). Adverbs and Functional Heads: A Cross-Linguistic Perspective. New York and Oxford: Oxford University Press. Clahsen, Harald (). Critical phases of grammar development: A study of the acquisition of negation in children and adults. In P. Jordens and J. Lalleman (eds.), Language Development, –. Dordrecht: Foris. Clahsen, Harald, and Pieter Muysken (). The availability of universal grammar to adult and child learners: A study of the acquisition of German word order. Second Language Research : –. Clahsen, Harald, Jürgen Meisel, and Manfred Pienemann (). Deutsch als Zweitsprache: Der Spracherwerb ausländischer Arbeiter. Tübingen: Narr. Clark, Eve V. (). The principle of contrast: A constraint on language acquisition. In B. macWhinney (ed.), Mechanisms of Language Acquisition, –. Hillsdale, NJ: Lawrence Erlbaum. Clark, H. H. (). Comprehending comparatives. In G. B. Flores d’Arcais and W. J. M. Levelt (eds.), Advances in Psycholinguistics, –. Amsterdam: North Holland Press. Clark, H. H. (). Semantics and comprehension. In T. A. Sebeok (ed.), Current Trends in Linguistics, Vol. : Linguistics and Adjacent Arts, –. The Hague: Mouton. Clark, H. H., and W. G. Chase (). On the process of comparing sentences against pictures. Cognitive Psychology  (): –. Clark, H. H., and W. G. Chase (). Perceptual coding strategies in the formation and verification of descriptions. Memory and Cognition : –. Clark, Ross (). Aspects of Proto-Polynesian Syntax. Auckland: Linguistic Society of New Zealand. Clifton, C., L. Frazier, and P. Deevy (). Feature manipulation in sentence comprehension. Italian Journal of Linguistics  (): –. Cohen, Ariel, and Manfred Krifka (). Superlative quantifiers as modifiers of meta-speech acts. The Baltic International Yearbook of Cognition, Logic and Communication : –. Coles-White, D’Jaris (). Negative concord in child African American English: Implications for specific language impairment. Journal of Speech, Language, and Hearing Research  (): –. Collier-Baker, Emma, and Thomas Suddendorf (). Do chimpanzees (Pan troglodytes) and -year-old children (Homo sapiens) understand double invisible displacement? Journal of Comparative Psychology : –. Collins, Chris (a). The distribution of negated quantifier phrases in English, posted at lingbuzz, . Collins, Chris (). Outer negation of universal quantifier phrases. Linguistics and Philosophy. . Collins, Chris, and Paul M. Postal (). Classical NEG Raising: An Essay in the Syntax of Negation. Cambridge, MA: MIT Press.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Collins, Chris, and Paul M. Postal (). Interclausal NEG raising and the scope of negation. Glossa  (): –. Collins, Chris, and Paul M. Postal (a). Dispelling the cloud of unknowing: More on the syntactic nature of Neg raising. In K. Turner and L. Horn (eds.), Pragmatics, Truth and Underspecification: Towards an Atlas of Meaning, –. Leiden: Brill. CRiSPI . Collins, Chris, and Paul M. Postal (b). Disentangling two distinct notions of NEG raising. Semantics and Pragmatics  (). (Early access version.) Collins, Chris, Paul M. Postal, and E. Yevudey (). Negative polarity items in Ewe. Journal of Linguistics. . Coltheart, M. (). Assumptions and methods in cognitive neuropsychology. In B. Rapp (ed.), Handbook of Cognitive Neuropsychology. New York: Psychology Press. Comrie, Bernard, Matthew S. Dryer, and Martin Haspelmath (). Introduction. In Dryer and Haspelmath (eds.) , accessed on January , . Comrie, Bernard, Martin Haspelmath, and Balthasar Bickel (). The Leipzig Glossing Rules. Max Planck Institute for Evolutionary Anthropology and University of Leipzig. . Condoravdi, Cleo (). Punctual until as a scalar NPI. In K. Hanson and S. Inkelas (eds.), The Nature of the Word: Studies in Honor of Paul Kiparsky, –. New York: The MIT Press. Condoravdi, Cleo, Dick Crouch, and Martin van den Berg (). Counting concepts. In R. van Rooy and M. Stokhof (eds.), Proceedings of the Thirteenth Amsterdam Colloquium, –. Amsterdam: ILLC. Condoravdi, Cleo, Dick Crouch, Martin van den Berg, John Everett, Valeria Paiva, Reinhard Stolle, and Danny Bobrow (). Preventing existence. Formal Ontology in Information Systems, ACM Press. Cooper, Robin (). Austinian propositions, Davidsonian events and perception complements. In J. Ginzburg, Z. Khasidashvili, J.-J. Levy, and E. Vallduvi (eds.), The Tbilisi Symposium on Language, Logic and Computaiton, –. Chicago: CSLI Publications. Cooper, Roger M. (). The control of eye fixation by the meaning of spoken language: A new methodology for the real-time investigation of speech perception, memory, and language processing. Cognitive Psychology : –. Coppock, Elizabeth, and David Beaver (). Principles of the exclusive muddle. Journal of Semantics : –. Coppock, Elizabeth, and David Beaver (). Definiteness and determinacy. Linguistics and Philosophy  (): –. Corblin, Francis (). Multiple negation processing. Human Communication Research Center. Paris (Report ). Corblin, Francis (). Compositionality and complexity in multiple negation. Logic Journal of IGPL : –. Corblin, Francis (). Multiple negation processing in natural language. Theoria : –. Corblin, Francis, and Lucia Tovena (). L’expression de la négation dans les langues romanes. In D. Godard (ed.), Les langues Romanes: Problèmes de la phrase simple, –. Paris: CNRS Editions. Corblin, Francis, Viviane Déprez, Henriette de Swart, and Lucia Tovena (). Negative concord. In Francis Corblin and Henriette de Swart (eds.), Handbook of French Semantics, –. Stanford, CA: CSLI. Corina, David P., Ursula Bellugi, and Judy Reilly (). Neuropsychological studies of linguistic and affective facial expressions in deaf signers. Language and Speech : –. Cormack, Annabel, and Neil Smith (). Modals and negation in English. In Sjef Barbiers, Frits Beukemam, and Wim van der Wurff (eds.), Modality and its Interaction with the Verbal System, –. Amsterdam: John Benjamins. Cornish, E. R. (). Pragmatic aspects of negation in sentence evaluation and completion tasks. British Journal of Psychology : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Cornulier, Benoît de (). Remarques à propos de la négation anticipée. Le français moderne : –. Corr, Alice V. (). Ibero-Romance and the Syntax of the Utterance. Ph.D. dissertation. University of Cambridge. Corver, Norbert (). Much-support as a last resort. Linguistic Inquiry : –. Coussi-Korbel, Sabine (). Learning to outwit a competitor in mangabeys (Cercocebus torquatus torquatus). Journal of Comparative Psychology : –. Crain, Stephen (). The Emergence of Meaning. Cambridge: Cambridge University Press. Crain, Stephen, and Rosalind Thornton (). Investigations in Universal Grammar: A Guide to Experiments on the Acquisition of Syntax and Semantics. Cambridge, MA: MIT Press. Crain, Stephen, and Rosalind Thornton (). rd year grammar. In A. Gallego and D. Ott (eds.),  Years Later: Reflections on Chomsky, –. Cambridge, MA: MIT Working Papers in Linguistics. Creissels, Denis, Gerrit J. Dimmendaal, Zygmunt Frajzyngier, and Crista König (). Africa as a morphosyntactic area. In Bernd Heine and Derek Nurse (eds.), A Linguistic Geography of Africa, –. Cambridge: Cambridge University Press. Crespo-Sendra, Verònica, Konstantin Kaland, Marc Swerts, and Pilar Prieto (). Perceiving incredulity: The role of intonation and facial gestures. Journal of Pragmatics : –. Cresti, Diana (). Extraction and reconstruction. Natural Language Semantics : –. Crnič, Luka (). Getting even. Ph.D. dissertation, MIT. Crockford, Catherine, Roman M. Wittig, and Klaus Zuberbühler (). Vocalizing in chimpanzees is influenced by social-cognitive processes. Science Advances : e. Crockford, Catherine, Roman M. Wittig, Roger Mundry, and Klaus Zuberbühler (). Wild chimpanzees inform ignorant group members of danger. Current Biology : –. Croft, William (). The evolution of negation. Journal of Linguistics  (): –. Crowley, Terry (). The Avava language of Central Malakula (Vanuatu). Canberra: Australian National University. Cruschina, Silvio (). Aspetti morfologici e sintattici degli avverbi in siciliano. In Jacopo Garzonio (ed.). Studi sui dialetti della Sicilia. Quaderni di lavoro ASIt , –. Padua: Unipress. Cruschina, Silvio (). The expression of evidentiality and epistemicity: Cases of grammaticalization in Italian and Sicilian. Probus : –. Cruse, D. Alan (). Antonyms and gradable complementaries. In D. Kastovsky (ed.), Perspektiven der lexicalischen Semantik, –. Bonn: Bouvier Verlag Herbert Grundmann. Crystal, Jonathon D., and Allison L. Foote (). Evaluating information-seeking approaches to metacognition. Current Zoology : –. Cunnings, I., and P. Sturt (). Coargumenthood and the processing of reflexives. Journal of Memory and Language : –. Cutler, Anne (). Lexical complexity and sentence processing. In G. B. Flores d’Arcais and R. J. Jarvella (eds.), The Process of Language Understanding, –. New York: Wiley. Cyffer, Norbert, Erwin Ebermann, and Georg Ziegelmeyer (eds.) (). Negation Patterns in West African Languages and Beyond. Amsterdam: Benjamins. Cysouw Michael (). The asymmetry of affixation. In H.-M. Gärtner, S. Beck, R. Eckardt, R. Musan, and B. Stiebels (eds.), Puzzles for Krifka, online publication, . Dahl, Östen (). Some notes on indefinites. Language : –. Dahl, Östen (). Typology of sentence negation. Linguistics : –. Dahl, Östen (). Inflationary effects in language and elsewhere. In Joan Bybee and Paul Hopper (eds.), Frequency and the Emergence of Linguistic Structure, –. Amsterdam and Philadelphia: Benjamins. Dahl, Östen (). Typology of negation. In Laurence R. Horn (ed.), The Expression of Negation, –. Berlin: de Gruyter.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Dale, R., and N. D. Duran (). The cognitive dynamics of negated sentence verification. Cognitive Science : –. Dally, Joanna M., Nathan J. Emery, and Nicola S. Clayton (). Food-caching western scrub-jays keep track of who was watching when. Science : –. Damasio, Antonio R. (). Aphasia. The New England Journal of Medicine  (): –. Damonte, Federico (). Modo e negazione nei dialetti calabresi meridionali. In Federica Cognola and Diego Pescarini (eds.), La negazione: Variazione dialettale ed evoluzione diacronica, –. Quaderni di lavoro ASIt . Padua: Unipress. Damonte, Federico (). Matching moods: Mood concord between CP and IP in Salentino and southern Calabrian subjunctive complements. In Paola Benincà and Nicola Munaro (eds.), Mapping the Left Periphery: The Cartography of Syntactic Structure, Vol. , –. Oxford/ New York: Oxford University Press. Davidson, Donald (). The logical form of action sentences. In N. Rescher (ed.), The Logic of Decision and Action. Pittsburgh: University of Pittsburgh Press. Davidson, Donald (). Causal relations. The Journal of Philosophy : –. Reprinted in D. Davidson, Essays on Actions and Events, –. Oxford: Clarendon Press, . Davies, Martin (). Acts and scenes. In Neil Cooper and Pascal Engel (eds.), New Inquiries into Meaning and Truth, –. New York: St. Martin’s Press. Davis, Wayne A. (). “Metalinguistic” negations, denial, and idioms. Journal of Pragmatics  (): –. Davison, A. (). Any as universal or existential? In J. van der Auwera (ed.), The Semantics of Determiners. London: Croom Helm. Dayal, Veneeta (). Locality in WH Quantification: Questions and Relative Clauses in Hindi. Dordrecht: Kluwer. Dayal, Veneeta (). Free relatives and ever: Identity and free choice readings. Proceedings of SALT : –. Dayal, Veneeta (). ANY as inherently modal. Linguistics and Philosophy : –. Dayal, Veneeta (). Single-pair versus multiple-pair answers: Wh-in-situ and scope. Linguistic Inquiry : –. Dayal, Veneeta (). The universal force of free choice any. In Linguistic Variation Yearbook, –. Amsterdam: John Benjamins Publishing Company. Deal, Amy Rose (). Modals without scales. Language : –. de Boysson-Bardies, B. D. (). Syntax and semantics in memorization of negation. In G. B. F. d’Arcais and W. J. M. Levelt (eds.), Advances in Psycholinguistics, –. Amsterdam: North-Holland. de Boysson-Bardies, B. D. (). On children’s interpretations of negation. Journal of Experimental Child Psychology : –. De Clercq, Karen (). Squat, zero and no/nothing: Syntactic negation vs. semantic negation. In Marion Elenbaas and Rick Nouwen (eds.), Linguistics in the Netherlands , –. Amsterdam: John Benjamins. De Clercq, Karen (). A unified syntax of negation. Ph.D. dissertation, Ghent University. De Clercq, Karen (a). The internal syntax of Q-words. Linguistics in the Netherlands : –. De Clercq, Karen (b). The nanosyntax of French negation: A diachronic perspective. In Silvio Cruschina, Katharina Hartmann, and Eva–Maria Remberger (eds.), Studies on Negation: Syntax, Semantics, and Variation, –. Göttingen: Vienna University Press. De Clercq, Karen (to appear). The morphosyntax of negative markers: A nanosyntactic account. Mouton de Gruyter. De Clercq, Karen (). Syncretisms and the morphosyntax of negation. In Lena Baunaz, Karen De Clercq, Liliane Haegeman, and Eric Lander (eds.), Exploring Nanosyntax, –. Oxford: Oxford University Press. De Clercq, Karen, and Guido Vanden Wyngaerd (). Why affixal negation is syntactic. In Aaron Kaplan et al. (eds.), Proceedings of the th West Coast Conference on Formal Linguistics, –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Somerville, MA: Cascadilla Proceedings Project . De Clercq, Karen and Guido Vanden Wyngaerd (a). Adjectives and negation: Deriving contrariety from contradiction. Papers of Linguistic Society of Belgium : –. De Clercq, Karen and Guido Vanden Wyngaerd (b). Negation and the functional sequence. Natural Language and Linguistic Theory  (): –. De Clercq, Karen, Liliane Haegeman, and Terje Lohndal (). Medial adjunct PPs in English: Implications for the syntax of sentential negation. Nordic Journal of Linguistics  (): –. De Cuypere, Ludovic (). Limiting the Iconic: From the Metatheoretical Foundations to the Creative Possibilities of Iconicity in Language. Amsterdam and Philadelphia: Benjamins. de Dios-Flores, I. (). Processing Multiple Negations: Grammatical Structures that are Perceived as Unacceptable. Talk at the th Biennial International Conference on the Linguistics of Contemporary English. Vigo, Spain. de Dios-Flores, I. H. Muller, and C. Phillips (). Negative Polarity Illusions: Licensors that Don’t Cause Illusions, and Blockers that Do. Poster at th Annual CUNY Conference on Human Sentence Processing, Massachusetts Institute of Technology. de Haan, Ferdinand (). The Interaction of Modality and Negation: A Typological Study. London and New York: Routledge. Dehaene, Stanislas (). The Number Sense. New York: Oxford University Press. Dehé, Nicole, and Bettina Braun (). The Prosody of Rhetorical Questions in English. Ms. University of Konstanz. De Houwer, Annick (). The Acquisition of Two Languages from Birth: A Case Study. Cambridge: Cambridge University Press. Delfitto, D., and G. Fiorin (). Negation in exclamatives. Studia Linguistica : –. Delfitto, D., C. Melloni, and M. Vender (). The (en)rich(ed) meaning of expletive negation. To appear in Evolutionary Linguistic Theory . Amsterdam: John Benjamins. Del Gobbo, Francesca, Nicola Munaro, and Cecilia Poletto (). On sentential particles: A crosslinguistic study. In S. Hancil, A. Haselow, and M. Post (eds.), Final Particles, –. Berlin: Mouton de Gruyter. Del Prete, Fabio (). The Semantics of the Italian Temporal Conjunctions Prima (before) and Dopo (after). Ph.D. dissertation, University of Milan. De Morgan, Augustus (). On the Syllogism, No. III, and on Logic in general. Transactions of the Cambridge Philosophical Society : –. den Besten, Hans (). Double negation and the genesis of Afrikaans. In Pieter Muysken and Norval Smith (eds.), Substrata versus Universals in Creole Languages, –. Dordrecht: Foris. den Dikken, Marcel, André Meinunger, and Chris Wilder (). Pseudoclefts and ellipsis. Studia Linguistica  (): –. Denizot, Camille (). Personne et rien dans les poèmes homériques: Emplois de οὔ τις et de οὐδείς. In Camille Denizot and Emmanuel Dupraz (eds.), Latin quis/qui, grec τις/τίς: Parcours et fonctionnements, –. Rouen: Presses universitaires de Rouen et du Havre. Depiante, Marcela Andrea (). The Syntax of Deep and Surface Anaphora: A Study of Null Complement Anaphora and Stripping/Bare Argument Ellipsis. University of Connecticut dissertation. Déprez, Viviane (). Two types of Negative Concord. Probus : –. Déprez, Viviane () A non unified analysis of negative concord. In Paul Hirschbulher and François Martineau (eds.), The Syntax and Semantics of Negation, –. Amsterdam: John Benjamin. Déprez, Viviane (). Parallel (a)symmetries and the internal structure of negative expression. Natural Language and Linguistic Theory : –. Déprez, Viviane (). Atoms of negation: An outside-in micro-parametric approach to negative concord. In Richard Ingham and Pierre Larrivée (eds.), The Evolution of Negation: Beyond the Jespersen Cycle, –. Berlin and New York: de Gruyter.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Déprez, Viviane (). What is strict negative concord: Lessons from French based Creoles. In Debra Ziegler and Bao Zhiming (eds.), Negation and Contact: With a Special Focus on Singlish, –. Amsterdam: John Benjamins. Déprez, Viviane, and F. Henri (). Negation and Negative Concord: The View from Creoles. Amsterdam: John Benjamins. Déprez, Viviane, and France Martineau (). Micro-parametric variation and Negative Concord. In Julie Auger, J. Clancy Clements, and Barbara Vance (eds.), Contemporary Approaches to Romance Linguistics: Selected Papers from the rd LSRL, –. Amsterdam: John Benjamins. Déprez, Viviane, and Amy Pierce (). Negation and functional projections in early grammar. Linguistic Inquiry : –. Déprez, Viviane, and Jeremy Yeaton (). French negative concord and discord: An experimental investigation of contextual and prosodic disambiguation. Proceedings of LSRL . Déprez, Viviane, Susagna Tubau, Anne Cheylus, and M.Teresa Espinal (). Double Negation in a Negative Concord language: An experimental investigation. Lingua : –. de Saussure, Ferdinand (). Cours de linguistique générale. Paris: Payot. Deschamps, Isabelle, Galit Agmon, Yonatan Loewenstein, and Yosef Grodzinsky (). The processing of polar quantifiers, and numerosity perception. Cognition, : –. de Swart, Henriëtte (). Meaning and use of not . . . until. Journal of Semantics  (): –. de Swart, Henriëtte (). Licensing of negative polarity items under inverse scope. Lingua : –. de Swart, Henriëtte (a). Negation, polarity and inverse scope. In H. Bennis, Martin B. H. Everaert, and Eric J. Reuland (eds.), Interface Strategies, –. Amsterdam: KNAW verhandelingen afdeling letterkunde. de Swart, Henriëtte (b). Scope ambiguities with negative quantifiers. In Klaus von Heusinger and Urs Egli (eds.), Reference and Anaphoric Relations. Dordrecht: Kluwer Academic Press. de Swart, Henriëtte (). Négation et coordination: La conjonction ni. In Reineke Bok-Bennema et al. (eds.), Adverbial Modification, –. Amsterdam: Rodopi. de Swart, Henriëtte (). Expression and Interpretation of Negation: An OT Typology. Dordrecht: Springer. de Swart, Henriëtte (). Negation. In Maria Aloni and Paul Dekker (eds.), The Cambridge Handbook of Formal Semantics, –. Cambridge: Cambridge University Press. de Swart, Henriëtte E., and Renata Fonville (). Double negatives and intonation in Dutch. In D. Gilbers and J. Hoeksema (eds.), Black Book: A Festchrift for Frans Zwarts, –. Gröningen: CLCG. de Swart, Henriëtte, and Ivan Sag (). Negation and negative concord in Romance. Linguistics and Philosophy : –. Detges, Ulrich, and Richard Waltereit (). Grammaticalisation vs. reanalysis: A semanticpragmatic account of functional change in grammar. Zeitschrift für Sprachwissenschaft : –. Deutsch, R., B. Gawronski, and F. Strack (). At the boundaries of automaticity: Negation as reflective operation. Journal of Personality and Social Psychology : –. Deutsch, R., R. Kordts-Freudinger, B. Gawronski, and F. Strack (). Fast and fragile: A new look at the automaticity of negation processing. Experimental Psychology : –. Deutscher, Max (). A note on saying and believing. Analysis : –. De Vega, Manuel, and Mabel Urrutia (). Counterfactual sentences activate embodied meaning. Journal of Cognitive Psychology : –. De Vega, Manuel, Mabel Urrutia, and Bernardo Riffo (). Cancelling updating in the comprehension of counterfactuals embedded in narratives. Memory and Cognition : –. De Vega, Manuel, Yurena Morera, Inmaculada León, David Beltrán, Pilar Casado, and Manuel Martín-Loeches (). Sentential negation might share neurophysiological mechanisms with action inhibition: Evidence from frontal theta rhythm. Journal of Neuroscience : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





De Villiers, J. G., and H. B. Tager Flusberg (). Some facts one simply cannot deny. Journal of Child Language : –. Devos, Maud, and Johan van der Auwera (). Jespersen Cycles in Bantu: Double and triple negation. In Journal of African Languages and Linguistics  (): –. Devos, Maud, and Daniel Van Olmen (). Describing and explaining the variation of Bantu imperatives and prohibitives. Studies in Language : –. Devos, Maud, Michael Kasombo Tshibanda, and Johan van der Auwera (). Jespersen cycles in Kanincin: Double, triple and maybe even quadruple negation. Africana Linguistica : –. de Vries, Lourens (). Some notes on the Tsaukambo language of West Papua. Language and Linguistics in Melanesia, Special issue I: –. Dickie, Imogen (). Negation, anti-realism, and the denial defense. Philosophical Studies  (): –. Diesing, M. (). Indefinites. Cambridge, MA: MIT Press. Dillon, B., A. Mishler, S. Sloggett, and C. Phillips (). Contrasting intrusion profiles for agreement and anaphora: Experimental and modeling evidence. Journal of Memory and Language  (): –. Dimmendaal, Gerrit Jan (). The Turkana Language. Dordrecht: Foris. Dimroth, Christine (). Age effects on the process of L acquisition? Evidence from the acquisition of negation and finiteness in L German. Language Learning  (): –. Dimroth, Christine (). The acquisition of negation. In Laurence R. Horn (ed.), The Expression of Negation, –. Berlin and New York: Mouton de Gruyter. Doetjes, Jenny (). The cameleontic nature of French ni: negative coordination in a negative concord language. Proceedings of Sinn und Bedeutung : –. Domaneschi, Fillipo, Maribel Romero, and Bettina Braun (). Bias in polar questions. Glossa: A Journal of General Linguistics  (), : –. Donaldson, M. (). Developmental aspects of performance with negatives. In G. B. F. d’Arcais and W. J. M. Levelt (eds.), Advances in Psycholinguistics, –. Amsterdam: North-Holland. Donati, C. (). A note on negation in comparison. Quaderni del Dipartimento di Linguistica : –. Università di Firenze. Doornenbal, Marius (). A Grammar of Bantawa: Grammar, Paradigm Tables, Glossary and Texts of a Rai Language of Eastern Nepal. Doctoral dissertation, Leiden University. Douglas, Jamie (). Māori subject extraction. Glossa: A Journal of General Linguistics  (): . –, . Drabbe, Peter (). Grammar of the Asmat Language. Syracuse, IN: Our Lady of the Lake Press. Drai, Dan, and Yosef Grodzinsky (a). A new empirical angle on the variability debate: Quantitative neurosyntactic analyses of a large data set from Broca’s Aphasia. Brain and Language  (): –. Drai, Dan, and Yosef Grodzinsky (b). The variability debate: More statistics, more linguistics. Brain and Language  (): –. Drenhaus, Heiner, Douglas Saddy, and Stefan Frisch (). Processing negative polarity items: When negation comes through the backdoor. In Stephan Kepser and Marga Reis (eds.), Linguistic Evidence (Studies in Generative Grammar [SGG] ), –. Berlin and New York: Mouton de Gruyter. Drozd, Kenneth (). Child English pre-sentential negation as metalinguistic exclamatory sentence negation. Journal of Child Language : –. Drozd, Kenneth (). Metalinguistic sentence negation in Child English. In Jack Hoeksema, Hotze Rullmann, Víctor Sanchez-Valencia, and Ton van der Wouden (eds.), Perspectives on Negation and Polarity Items, –. Amsterdam/Philadelphia: John Benjamins. Dryer, Matthew S. (). Universals of negative position. In Michael Hammond, Edith A. Moravcsik, and Jessica Wirth (eds.), Studies in Syntactic Typology, –. Amsterdam: Benjamins.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Dryer, Matthew S. (). Focus, pragmatic presupposition, and activated propositions. Journal of Pragmatics : –. Dryer, Matthew S. (). Verb-object-negative order in Central Africa. In Cyffer et al. (eds.), Negation Patterns in West African Languages and Beyond, –. Dryer, Matthew S. (). Negative morphemes. In Matthew S. Dryer and Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Munich: Max Planck Digital Library. . Dryer, Matthew S. (a). Negative morphemes. In Dryer and Haspelmath (eds.), . Dryer, Matthew S. (b). Order of negative morpheme and verb. In Dryer and Haspelmath (eds.), . Dryer, Matthew S. (c). Position of negative morpheme with respect to subject, object, and verb. In Dryer and Haspelmath (eds.), . Dryer, Matthew S., and Martin Haspelmath (eds.) (). The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at ). Du, P., D. Liu, L. Zhang, G. Hitchman, and C. Lin (). The processing of contradictory and noncontradictory negative sentences. Journal of Cognitive Psychology : –. Ducrot, Oswald (). Dire et ne pas dire: Principes de sémantique linguistique. Paris: Hermann. Ducrot, Oswald (). La preuve et le dire: Langage et logique. Paris: Mame. Ducrot, Oswald (). Les échelles argumentatives. Paris: Minuit. Dudschig, C., and B. Kaup (). How does “not left” become “right”? Electrophysiological evidence for a dynamic conflict-bound negation processing account. Journal of Experimental Psychology: Human Perception and Performance  (): –. Dudschig, C., I. de la Vega, and B. Kaup (). To fly or not to fly? The automatic influence of negation on language–space associations. Cognitive Processing  (Suppl. ): –. Dudschig, C., I. G. Mackenzie, H. Leuthold, and B. Kaup (). Environmental sound priming: Does negation modify N cross-modal priming effects? Psychonomic Bulletin and Review : –. Dudschig, C., I. G. Mackenzie, C. Maienborn, B. Kaup, and H. Leuthold (). Negation and the N: Investigating temporal aspects of negation integration using semantic and world-knowledge violations. Language, Cognition and Neuroscience  (): –. Duffield, Nigel (). Particles and Projections in Irish Syntax. Boston: Kluwer. Duffield, Nigel, and Lydia White (). Assessing L knowledge of Spanish clitic placement: Converging methodologies. Second Language Research  (): –. Duffley, Patrick, and Pierre Larrivée (). Need, dare, and negative polarity. Linguistic Analysis : –. Duffley, Patrick, and Pierre Larrivée (). Anyone for non-scalarity? English Language and Linguistics  (): –. Dummet, Michael (). Frege: Philosophy of Language. London: Duckworth. Dummett, Michael (). The Logical Basis of Metaphysics. London: Duckworth. Dwivedi, V. D. (). Negation as a functional projection in Hindi. In K. Hunt, T. A. Perry, and V. Samiian (eds.), Proceedings of the Western Conference on Linguistics, –. Fresno, CA: Department of Linguistics, California State University of Fresno. Dwivedi, V. D. (). Modality and discourse processing. In S. Montrul and M. Kessler Robb (eds.), McGill Working Papers in Linguistics, : –. Montréal: Department of Linguistics, McGill University. Dwivedi, V. D. (). Interpreting quantifier scope ambiguity: Evidence of heuristic first, algorithmic second processing. PLoS ONE  (): –. Dwivedi, V. D. (). Individual Differences in Sentence Interpretation. Invited presentation for the TESL Niagara Spring Conference, Niagara College, Niagara-on-the-Lake, Canada.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Dwivedi, V. D. (). The role of affect in sentence perception. Poster presentation for st annual meeting of the Cognitive Science Society, Montreal, Canada Dwivedi, V. D., and M. Goldhawk (). Underspecification of Scope Ambiguity and Anaphora: Evidence from Self-Paced Reading. Poster presented at the nd annual meeting of the CUNY Sentence Processing Conference, Davis, CA. Dwivedi, V. D., and J. Selvanayagam (). N, dispositional affect and sentence processing. Poster presentation for th annual meeting of the Cognitive Neuroscience Society, San Francisco, USA. Dwivedi, V. D., K. E. Goertz, and J. Selvanayagam (). Heuristics in language comprehension. Journal of Behavioural and Brain Sciences  (): –. Dwivedi, V. D., L. Rowland, and K. Curtiss (). Effects of Semantic Context on Heuristic vs. Algorithmic Processing. Poster session presented at the nd annual meeting of the Cognitive Neuroscience Society, San Francisco. Dwivedi, V. D., N. A. Phillips, M. Laguë-Beauvais, and S. R. Baum (). An electrophysiological study of mood, modal context, and anaphora. Brain Research  (): –. Dwivedi, V. D., J. E. Drury, M. Molnar, N. A. Phillips, S. Baum, and K. Steinhauer (a). ERPs reveal sensitivity to hypothetical contexts in spoken discourse. NeuroReport  (): –. Dwivedi, V. D., N. A. Phillips, S. Einagel, and S. R. Baum (b). The neural underpinnings of semantic ambiguity and anaphora. Brain Research : –. Dwyer, Dominic M., and Katy V. Burgess (). Rational accounts of animal behaviour? Lessons from C. Lloyd Morgan’s canon. International Journal of Comparative Psychology : –. Dwyer, Dominic M., and Michael R. Waldmann (). Beyond the information (not) given: Representations of stimulus absence in rats (Rattus norvegicus). Journal of Comparative Psychology : –. Early, Robert (a). A Grammar of Lewo, Vanuatu. Canberra: The Australian National University Ph.D. dissertation. Early, Robert (b). Lewo. In Kahrel and van den Berg (eds.), Typological Studies in Negation, –. Eckardt, Regine (). Meaning Change in Grammaticalization: An Enquiry into Semantic Reanalysis. Oxford: Oxford University Press. Eckardt, Regine (). The lower part of even ontology. In Johannes Dölling, Tatjana HeydeZybatow, and Martin Schäfer (eds.), Event Structures in Linguistic Form and Interpretation, –. Berlin and New York: Walter de Gruyter. Eckardt, Regine (). The many careers of negative polarity items. In Kristin Davidse, Tine Breban, Lotte Brems, and Tanja Mortelmans (eds.), Grammaticalization and Language Change: New Reflections, –. Amsterdam: Benjamins. Eckardt, Regine, and Eva Csipak (). Minimizers—towards pragmatic licensing. In Eva Csipak, Regine Eckardt, Mingya Liu, and Manfred Sailer (eds.), Beyond “ever” and “any”: New Explorations in Negative Polarity Sensitivity, –. Berlin: de Gruyter. Edmondson, Jerry A. (). Polarized auxiliaries. In Frank Heny and Barry Richards (eds.), Linguistic Categories: Auxiliaries and Related Puzzles, Vol. : Categories, –. Dordrecht: Reidel. Edmiston, Daniel (). Negative concord in an English of the Southern US and beyond. Florida Linguistic Papers –. Ehala, Martin (). Self-organisation and language change. Diachronica : –. Eiferman, R. R. (). Negation, a linguistic variable. Acta Psychologica : –. Eilam, Aviad (). Expletive negation in modern Hebrew: Neither expletive nor negation. Talk presented at the rd Annual Meeting of the Israel Association for Theoretical Linguistics (IATL), June –, . Elbert, Samuel H. (). Spoken Hawai’an. Honolulu: University of Hawai’i Press. Ellis, Rod (). The origins of interlanguage. Applied Linguistics  (): –. Ellis, Rod (). Researching acquisition sequences: Idealization and de-idealization in SLA. Language Learning  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Elšik, Viktor, and Yaron Matras (). Markedness & Language Change. Berlin: Mouton de Gruyter. Emonds, Joseph E. (). The verbal complex V0 -V in French. Linguistic Inquiry : –. Engdahl, Elisabet (). Constituent Questions. Dordrecht: D. Reidel Publishing Company. Engelkamp, J., and H. Hörmann (). The effect of non-verbal information on the recall of negation. Quarterly Journal of Experimental Psychology : –. Engelkamp, J., F. Merdian, and H. Hörmann (). Semantische Faktoren beim Behalten der Verneinung von Sätzen. Psychologische Forschung : –. England, Nora (). A Grammar of Mam, a Mayan Language. Austin: University of Texas Press. Epstein, Samuel, Suzanne Flynn, and Gita Martohardjono (). Second language acquisition: Theoretical and experimental issues in contemporary research. Behavioral and Brain Sciences  (): –. Erlewine, Michael Yoshitaka (). Movement out of Focus. Ph.D. dissertation, MIT. Ernst, Thomas (). Adverbs and positive polarity in Mandarin Chinese. In Marjorie K. M. Chan and Hana Kang (eds.), Proceedings of the th North American Conference on Chinese Linguistics, –. Ohio: The Ohio State University. Erotokritou, Vasiliki (). The Semantics and Pragmatics of Tag Questions in Cypriot Greek. MA thesis, University of Konstanz. Escandell-Vidal, Victoria (). Los enunciados interrogativos: Aspectos semánticos y pragmáticos. In I. Bosque and V. Demonte (eds.), Gramática descriptiva de la lengua española, Vol. , –. Madrid: Espasa Calpe. Espinal, M.Teresa (). Expletive negation and logical absorption. The Linguistic Review  (): –. Espinal, M.Teresa (). Non-negative negation and wh-exclamatives. In D. Forget, P. Hirschbühler, F. Martineau, and M. L. Rivero (eds.), Negation and Polarity: Syntax and Semantics. Amsterdam: John Benjamins. Espinal, M.Teresa (). On the semantic status of n-words in Catalan and Spanish. Lingua : –. Espinal, M.Teresa (). Licensing expletive negation and negative concord in Catalan and Spanish. In F. Floricic (ed.), La négation dans les langues romanes. Linguisticae Investigationes Supplementa . Amsterdam: John Benjamins. Espinal, M.Teresa (). Pragmatic particles at the syntax-cognition interface. In L. Payrató and J. M. Cots (eds.), The Pragmatics of Catalan, –. Berlin: Mouton Series in Pragmatics, MSP . Espinal, M.Teresa, and Pilar Prieto (). Intonational encoding of double negation in Catalan. Journal of Pragmatics : –. Espinal, M.Teresa, and Susagna Tubau (a). Interpreting argumental n-words as answers to negative wh-questions. Lingua : –. Espinal, M.Teresa, and Susagna Tubau (b). Meaning of words, meaning of sentences: Building the meaning of n-words. In Susann Fischer and Christoph Gabriel (eds.), Grammatical Interfaces in Romance Linguistics, –. Berlin: Mouton de Gruyter. Espinal, M.Teresa, and Susagna Tubau (). Response systems: The syntax and semantics of fragment answers and response particles. Annual Review of Linguistics : –. . Espinal, M.Teresa, Susagna Tubau, Joan Borràs-Comes, and Pilar Prieto (). Double negation in Catalan and Spanish: Interaction between syntax and prosody. In Pierre Larrivée and Chungmin Lee (eds.), Negation and Polarity: Experimental Perspectives, –. Cham: Springer. Etxeberria, Urtzi, Susagna Tubau, Viviane Déprez, Joan Borràs-Comes, and M.Teresa Espinal (). Relating (un)acceptability to interpretation: Experimental investigations on negation. Frontiers in Psychology . Article . Etxepare, Ricardo (). Negation. In José Ignacio Hualde and Jon Ortiz de Urbina (eds.), A Grammar of Basque, –. Berlin and New York: Mouton de Gruyter.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Etxepare, Ricardo, and Myriam Uribe-Etxebarria (). Negation and focus in Spanish and Basque. Supplements of the International Journal of Basque Language and Linguistics : –. Eubank, Lynn (/). On the transfer of parametric values in L development. Language Acquisition : –. Eubank, Lynn (). Negation in early German–English Interlanguage: More Valueless Features in the L initial state. Second Language Research  (): –. Everett, Caleb (). Patterns in Karitiana: Articulation, perception, and grammar. Doctoral dissertation, Rice University. Fălăuş, Anamaria (). Negative concord and double negation: The Romanian puzzle. In J. Camacho, N. Flores-Ferran, L. Sanchez, V. Déprez, and M. J. Cabrera (eds.), Romance Linguistics , –. Amsterdam: John Benjamins. Fălăuş, Anamaria (a). Introduction: Alternatives in semantics and pragmatics. In A. Fălăuş (ed.), Alternatives in Semantics, –. Basingstoke: Palgrave Macmillan. Fălăuş, Anamaria (b). Broaden your views, but try to stay focused: A missing piece in the polarity system. In I. Caponigro and C. Cecchetto (eds.), From Grammar to Meaning: The Spontaneous Logicality of Language, –. Cambridge: Cambridge University Press. Fălăuş, Anamaria, and Andreea C. Nicolae (). Fragment answers and double negation in strict negative concord languages. In Mary Moroney, Carol-Rose Little, Jacob Collard, and Dan Burgdoff (eds.), Proceedings of SALT : –. Ithaca, NY: Linguistic Society of America and Cornell Linguistics Circle. Farkas, Donka F., and Kim B. Bruce (). On reacting to assertions and polar questions. Journal of Semantics  (): –. Farkas, Donka F., and Anastasia Giannakidou (). How clause-bounded is the scope of universals? In Semantics and Linguistic Theory (SALT) : –. Ithaca, NY: Cornell University. Farkas, Donka F., and Floris Roelofsen (). Division of labor in the interpretation of declaratives and interrogatives. Journal of Semantics  (): –. Fassi-Fehri, Abdelkader (). Issues in the Structure of Arabic Clauses and Words. Dordrecht: Kluwer. Fast, Cynthia D., and Aaron P. Blaisdell (). Rats are sensitive to ambiguity. Psychonomic Bulletin & Review : –. Fast, Cynthia D., Traci Biedermann, and Aaron P. Blaisdell (). Imagine that! Cue-evoked representations guide rat behavior during ambiguous situations. Journal of Experimental Psychology: Animal Learning and Cognition : –. Fauconnier, Gilles (a). Polarity and the scale principle. Chicago Linguistic Society (CLS) : –. Fauconnier, Gilles (b). Pragmatic scales and logical structure. Linguistic Inquiry  (): –. Fauconnier, Gilles (). Implicational reversal in a natural language. In Franz Guenthner and S. J. Schmidt (eds.), Formal Semantics and Pragmatics for Natural Languages, –. Dordrecht: Reidel. Fauconnier, Gilles (). Mental Spaces: Aspects of Meaning Construction in Natural Language. Cambridge, MA: MIT Press. Feagin, Crawford (). Variation and Change in Alabama English. Washington, DC: Georgetown University Press. Feiman, R., S. Carey, and F. Cushman (). Infants’ representations of others’ goals: Representing approach over avoidance. Cognition : –. Feiman, R., S. Mody, S. Sanborn, and S. Carey (). What do you mean, no? Toddlers’ comprehension of logical ‘no’ and ‘not’. Language Learning and Development  (): –. Ferguson, H. J., and A. J. Sanford (). Anomalies in real and counterfactual worlds: An eyemovement investigation. Journal of Memory and Language : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Ferguson, H. J., A. J. Sanford, and H. Leuthold (). Eye-movements and ERPs reveal the time course of processing negation and remitting counterfactual worlds. Brain Research : –. Fernald, Anne (). Maternal vocalisations to infants as biologically relevant signals: An evolutionary perspective. In J. H. Barkow, L. Cosmides, and J. Tooby (eds.), The Adapted Mind: Evolutionary Psychology and the Generation of Culture. Oxford: Oxford University Press. Ferreira, F. (). The misinterpretation of noncanonical sentences. Cognitive Psychology  (): –. Ferreira, F., and C. Clifton (). The independence of syntactic processing. Journal of Memory and Language  (): –. Ferreira, F., K. G. D. Bailey, and V. Ferraro (). Good-enough representations in language comprehension. Current Directions in Psychological Science,  (): –. Ferry, A. L., S. J. Hespos, and D. Gentner (). Prelinguistic relational concepts: Investigating analogical processing in infants. Child Development  (): –. Field, Hartry (). Saving Truth from Paradox. Oxford: Oxford University Press. Fiengo, Robert, and Robert May (). Indices and Identity. Cambridge, MA: MIT Press. Filiaci, Francesca, Antonella Sorace, and Manuel Carreiras (). Anaphoric biases of null and overt subjects in Italian and Spanish: A cross-linguistic comparison. Language, Cognition and Neuroscience  (): –. Filik, R., K. B. Paterson, and S. P. Liversedge (). Processing doubly quantified sentences: Evidence from eye movements. Psychonomic Bulletin and Review,  (): –. Fillenbaum, Samuel (). Memory for Gist: Some relevant variables. Language and Speech : –. Fillmore, Charles J. (). The position of embedding transformations in a grammar. Word : –. Fillmore, Charles J. (). Types of lexical information. In F. Kiefer (ed.), Studies in Syntax and Semantics, –. Dordrecht: Reidel. Fischer, Susan (). Questions and negation in American Sign Language. In Ulrike Zeshan (ed.), Interrogative and Negative Constructions in Sign languages, –. Nijmegen: Ishara Press. Fischler, I., P. A. Bloom, D. G. Childers, S. E. Roucos, and N. W. Perry Jr (). Brain potentials related to stages of sentence verification. Psychophysiology  (): –. Fisher, John (). The Emergence of Standard English. Kentucky: University Press of Kentucky. Fleisher, Nicholas (). Comparative quantifiers and negation: Implications for Scope Economy. Journal of Semantics : –. Flick, Graham, Yohei Oseki, Amanda R. Kaczmarek, Meera Al Kaabi, Alec Marantz, and Liina Pylkkänen (). Building words and phrases in the left temporal lobe. Cortex, : –. Flombaum, Jonathan I., and Laurie R. Santos (). Rhesus monkeys attribute perceptions to others. Current Biology : –. Fodor, Janet D. (). The Linguistic Description of Opaque Contexts. Ph.D. dissertation, MIT. Fodor, Janet D. (). The mental representaion of quantifiers. In S. Peters and E. Saarinen (eds.), Processes, Beliefs, and Questions: Essays on Formal Semantics of Natural Language and Natural Language Processing, –. Dordrecht: Reidel. Fodor, Janet D., Jerry A. Fodor, and Merrill F. Garrett (). The psychological unreality of semantic representations. Linguistic Inquiry : –. Fodor, Jerry A., and Merrill F. Garrett (). Some reflections on competence and performance. Psycholinguistic Papers –. Fonseca-Greber, Bonnibeth Beale (). The emergence of emphatic ‘ne’ in Conversational Swiss French. Journal of French Language Studies  (): –. Fonville, Renata (). The Role of Intonation in the Use of Double Negatives in Dutch. MA thesis, Utrecht University.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Fonville, Renata, and Henriëtte de Swart (). Double negatives and intonation in Dutch. In Dicky Gilbers and Jack Hoeksema (eds.), Black Book: A Festschrift for Frans Zwarts, –. Groningen: CLCG working papers in Linguistics, Groningen University. Foolen, Ad (). Metalinguistic negation and pragmatic ambiguity: Some comments on a proposal by Laurence Horn. Pragmatics  (): –. Foreman, John (). Syntax of negative inversion in non-standard English. In Kimary Shahin, Susan Blake, and Eun-Sook Kim (eds.), Proceedings of WCCFL . Stanford, CA: CSLI. Foroni, Francesco, and Gün R. Semin (). Comprehension of action negation involves inhibitory simulation. Frontiers in Human Neuroscience : . Fox, Danny (). Focus, Parallelism and accommodation. In Tanya Matthews and Devon Strolovitch (eds.), Proceedings of SALT : –. Ithaca, NY: Cornell University. Fox, Danny (). Economy and Semantic Interpretation. Cambridge, MA: MIT Press. Fox, Danny (). Antecedent-contained deletion and the copy theory of movement. Linguistic Inquiry : –. Fox, Danny (a). Free choice disjunction and the theory of scalar implicatures. In U. Sauerland and P. Stateva (eds.), Presupposition and Implicature in Compositional Semantics, –. London: Palgrave Macmillan. Fox, Danny (b). Too many alternatives: Density, symmetry and other predicaments. In T. Friedman and M. Gibson (eds.), Proceedings of SALT XVII, –. Ithaca, NY: Cornell University. Fox, Danny, and Martin Hackl (). The universal density of measurement. Linguistics and Philosophy : –. Fox, Danny, and Roni Katzir (). On the characterization of alternatives. Natural Language Semantics : –. Fox, Danny, and David Pesetsky (). Cyclic linearization of syntactic structure. Theoretical Linguistics  (): –. Fox, Danny, and Benjamin Spector (). Economy and embedded exhaustification. Natural Language Semantics : –. Francis, Naomi (). Modal scope in negative inversion constructions. In Aaron Kaplan, Abby Kaplan, Miranda K. McCarvel, and Edward J. Rubin (eds.), Proceedings of the th West Coast Conference on Formal Linguistics, –. Somerville, MA: Cascadilla Proceedings Project. Francis, Naomi (). PPI modals and scope in negative inversion constructions. Ms. MIT. Francis, Winthrop N., and Henry Kucera (). Frequency Analysis of English Usage: Lexicon and Grammar. Boston: Houghton Mifflin. Franklin, Amy, Anastasia Giannakidou, and Susan Goldin-Meadow (). Negation, questions, and structure building in a home sign system. Cognition  (): –. Frazier, L., and J. D. Fodor (). The sausage machine: A new two-stage parsing model. Cognition  (): –. Frazier, L., and K. Rayner (). Making and correcting errors during sentence comprehension: Eye movements in the analysis of structurally ambiguous sentences. Cognitive Psychology  (): –. Frege, Gottlob (). Über Sinn und Bedeutung. Zeitschrift für Philosophie und philosophische Kritik : –. Frege, Gottlob (/). Logic. In Posthumous Writings, ed. Hans Hermes, Friedrich Kambartel, and Friedrich Kaulbach, –. Oxford: Basil Blackwell. Frege, Gottlob (). Negation. In M. Beaney (ed.), The Frege Reader, –. Oxford: Blackwell. Frege, Gottlob (/). Negation. In Translations from the Philosophical Writings of Gottlob Frege, ed. Peter Geach and Max Black, –. Oxford: Oxford University Press. French, Rohan (). Structural reflexivity and the paradoxes of self-reference. Ergo  (): –. Friederici, Angela D. (). Towards a neural basis of auditory sentence processing. Trends in Cognitive Sciences  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Friederici, Angela D. (). The brain basis of language processing: From structure to function. Physiological Reviews  (): –. Friederici, Angela D., and Jürgen Weissenborn (). Mapping sentence form onto meaning: The syntax–semantics interface. Brain Research : –. Friedmann, Na’ama (). Agrammatism and the psychological reality of the syntactic tree. Journal of Psycholinguistic Research  (): –. Friedmann, Na’ama (). Degrees of severity and recovery in agrammatism: Climbing up the syntactic tree. Aphasiology  (–): –. Frisch, Stefan, Matthias Schlesewsky, Douglas Saddy, and Annegret Alpermann (). The P as an indicator of syntactic ambiguity. Cognition  (): B–B. Fruehwald, Josef, Jonathan Gress-Wright, and Joel C. Wallenberg (). Phonological rule change: The constant rate effect. In Proceedings of th Meeting of the Northeast Linguistic Society (NELS) . Fuster, Joaquin M. (). Cortex and Mind: Unifying Cognition. Oxford: Oxford University Press. Fyndanis, Valantis, Kyrana Tsapkini, Spyridoula Varlokosta, Konstantina Petropoulou, and Ilias Papathanasiou (). Negation in agrammatism: Evidence from Greek. Brain and Language  (–): –. Gajewski, Jon (). L-analyticity in natural language. Ms. MIT. Gajewski, Jon (). Neg-Raising: Polarity and Presupposition. Ph.D. dissertation, MIT. Gajewski, Jon (). Neg-raising and polarity. Linguistics and Philosophy : –. Gajewski, Jon (). More on quantifiers in comparative clauses. In Tova Friedman and Satoshi Ito (eds.), Proceedings of Semantics and Linguistic Theory : –. Gajewski, Jon (). Licensing strong NPIs. Natural Language Semantics  (): –. García-Marco, E., Y. Morera, D. Beltrán, M. de Vega, E. Herrera, L. Sedeño, A. Ibáñez, and A. M. García (). Negation markers inhibit motor routines during typing of manual action verbs. Cognition : –. García Negroni, María Marta (). Metalinguistic negation and rejection discourse markers in Spanish. In Malin Roitman (ed.), The Pragmatics of Negation: Negative Meanings, Uses and Discursive Functions, –. Amsterdam/Philadelphia: John Benjamins. Gardiner, Alan H. (). The word iwn. Zeitschrift für Ägyptische Sprache und Altertumskunde : –. Garzonio, Jacopo (). A case of incomplete Jespersen’s cycle in Romance. Rivista di Grammatica Generativa : –. Garzonio, Jacopo, and Cecilia Poletto (). The negative marker that escaped the cycle: Some notes on manco. In Carla Contemori and Lena Dal Pozzo (eds.), Inquiries into Linguistic Theory and Language Acquisition: Papers Offered to Adriana Belletti, –. Siena: CISCL Press. Garzonio, Jacopo, and Cecilia Poletto (a). On polarity particles in Italian varieties. In Josef Bayer, Roland Hinterhölzl, and Andreas Trotzke (eds.), Discourse-Oriented Syntax, –. Amsterdam: Benjamins. Garzonio, Jacopo, and Cecilia Poletto (b). On preverbal negation in Sicilian and syntactic parasitism. Isogloss: A Journal on Variation of Romance and Iberian Languages special issue: –. Gast, Volker, and Johan van der Auwera (). Scalar additive operators in the languages of Europe. Language  (): –. Gauker, Christopher (). What is a context of utterance? Philosophical Studies  (): –. Gawronski, B., R. Deutsch, S. Mbirkou, B. Seibt, and F. Strack (). When “Just Say No” is not enough: Affirmation versus negation training and the reduction of automatic stereotype activation. Journal of Experimental Social Psychology : –. Gazdar, Gerald (a). Pragmatics: Implicature, Presupposition, and Logical Form. New York: Academic Press. Gazdar, Gerald (b). A solution to the projection problem. In C. Oh and D. Dinneen (eds.), Syntax and Semantics, Vol. : Presupposition, –. New York: Academic Press.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Geach, P. T. (). Assertion. The Philosophical Review  (): –. Geach, P. T. (). Logic Matters. Berkeley: University of California Press. Geçkin, Vasfiye, Rosalind Thornton, and Stephen Crain (). Children’s interpretation of disjunction in negative sentences: A comparison of Turkish and German. Language Acquisition  (): –. Gengel, Kirsten (). Pseudogapping and Ellipsis. Oxford: Oxford University Press. George, Ken (). Cornish. In Martin Ball and Nicole Müller (eds.), The Celtic Languages, –. New York: Routledge. Geraci, Carlo (). Negation in LIS (Italian Sign Language). In Leah Bateman and Cheron Ussery (eds.), Proceedings of the North East Linguistic Society (NELS ). Amherst, MA: GLSA. Gernsbacher, Morton Ann (). Language Comprehension as Structure Building. Hillsdale, NJ: Erlbaum. Geurts, Bart (). The mechanisms of denial. Language  (): –. Geurts, Bart (a). On no. Journal of Semantics : –. Geurts, Bart (b). Presuppositions and Pronouns. Oxford: Elsevier. Geurts, Bart, and Frans Van Der Slik (). Monotonicity and processing load. Journal of Semantics  (): –. Giannakidou, Anastasia (). The Landscape of Polarity Items. Ph.D. dissertation. University of Groningen. Giannakidou, Anastasia (). Polarity Sensitivity as (Non)veridical Dependency. Amsterdam: John Benjamins. Giannakidou, Anastasia (). Affective dependencies. Linguistics and Philosophy : –. Giannakidou, Anastasia (a). Negative . . . concord? Natural Language and Linguistic Theory : –. Giannakidou, Anastasia (b). Negative concord and the scope of universals. Transactions of the Philological Society  (): –. Giannakidou, Anastasia (). The meaning of free choice. Linguistics and Philosophy : –. Giannakidou, Anastasia (). UNTIL, aspect and negation: A novel argument for two untils. In Proceedings of Semantics and Linguistic Theory XII, –. Giannakidou, Anastasia (a). N-words and negative concord. In Martin Everaert and Henk van Riemsdijk (eds.), The Blackwell Companion to Syntax, Vol. , –. Malden, MA: Blackwell. Giannakidou, Anastasia (b). Only, emotive factive verbs, and the dual nature of polarity dependency. Language  (): –. Giannakidou, Anastasia (). The landscape of EVEN. Natural Language and Linguistic Theory  (): –. Giannakidou, Anastasia (). The dependency of the subjunctive revisited: Temporal semantics and polarity. Lingua  (): –. Giannakidou, Anastasia (). The dynamic of change in Dutch enig: From nonveridicality to strong negative polarity. Natural Language and Linguistic Theory : –. Giannakidou, Anastasia (). Positive polarity items and negative polarity items: Variation, licensing, and compositionality. In Claudia Maienborn, Klaus von Heusinger, and Paul Portner (eds.), Semantics: An International Handbook of Natural Language Meaning, –. Berlin: de Gruyter. Giannakidou, Anastasia (). Veridicality in Grammar and Thought: Modality, Propositional Attitudes and Negation. Chicago: University of Chicago Press. Giannakidou, Anastasia (). A critical assessment of exhaustivity for negative polarity items: The view from Greek, Korean, Mandarin, and English. Acta Academic Linguistica : –. Giannakidou, Anastasia, and Lisa Cheng (). (In)definiteness, polarity, and the role of whmorphology in free choice. Journal of Semantics : –. Giannakidou, Anastasia, and Urtzi Etxeberria (). Assessing the role of experimental evidence for interface judgment: Licensing of negative polarity items, scalar readings and focus. Frontiers in Psychology : .

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Giannakidou, Anastasia and Alda Mari (). The semantic roots of positive polarity: Epistemic modal verbs and adverbs in English, Greek and Italian. Linguistics and Philosophy (): –. Giannakidou, Anastasia, and Josep Quer (). Two mechanisms for the licensing of negative indefinites. In Leslie Gabriele, Debra Hardison, and Robert Westmoreland (eds.), FLSM VI: Proceedings of the th Annual Meeting of the Formal Linguistics Society of Mid-America, Vol. , –. Bloomington: Indiana University. Giannakidou, Anastasia, and Josep Quer (). Long-distance licensing of negative indefinites. In Danielle Forget, Paul Hirschbühlter, France Martineu, and María Luisa Rivero (eds.), Negation and Polarity: Syntax and Semantics, –. Amsterdam: John Benjamins. Giannakidou, Anastasia, and Josep Quer (). Exhaustive and non-exhaustive variation with free choice and referential vagueness: Evidence from Greek, Catalan, and Spanish. Lingua : –. Giannakidou, Anastasia, and Melita Stravou (). On metalinguistic comparatives and negation in Greek. In Claire Halpert, Jeremy Hartman, and David Hill (eds.), Proceedings of the  Workshop on Greek Syntax and Semantics, MIT Working Papers in Linguistics : –. Giannakidou, Anastasia, and Suwon Yoon (). The subjective mode of comparison: Metalinguistic comparatives in Greek and Korean. Natural Language & Linguistic Theory  (): –. Giannakidou, Anastasia, and Suwon Yoon (). Scalar marking without scalar meaning: Nonscalar, non-exhaustive NPIs in Greek and Korean. Language : –. Giannakidou, Anastasia, and Hedde Zeijlstra (). The landscape of negative dependencies: Negative concord, N-words. In M. Everaert and H. van Riemsdijk (eds.), The Blackwell Companion to Syntax, nd edn. London, Blackwell. Giannakidou, Anastasia, and Frans Zwarts (). Aspectual properties of temporal connectives. In Greek Linguistics ’: Proceedings of the rd International Conference on Greek Linguistics, –. Athens: Ellinika Grammata. Gianollo, Chiara (). Indefinites between Latin and Romance. Habilitation dissertation, Köln Universität. Gianollo, Chiara (). Focus-sensitive negation in Latin. Catalan Journal of Linguistics : –. Gianollo, Chiara (). Indefinites between Latin and Romance. Oxford: Oxford University Press. Gibson, E. A., E. Fedorenko, S. Piantadosi, and A. Gualmini (). Inter-subject variability in the ability to use context during language comprehension. Poster session presented at the th annual meeting of the CUNY Sentence Processing Conference, New York. Gibson, E. J. (). Exploratory behavior in the development of perceiving, acting, and the acquiring of knowledge. Annual Review of Psychology  (): –. Gibson, E. J., and R. D. Walk (). The “visual cliff ”. Scientific American  (): –. Gigerenzer, Gerd (). Adaptive Thinking: Rationality in the Real World. New York: Oxford University Press. Gilkerson, Jill, Nina Hyams, and Susan Curtiss (). On the scope of negation: More evidence for early parameter setting. Proceedings of GALA . University of Utrecht LOT. Ginzburg, Jonathan, and Ivan Sag (). Interrogative Investigations. Stanford, CA: CSLI Publications. Giora, R., D. Zimmerman, and O. Fein (). How can you compare! On negated comparisons as comparisons. Intercultural Pragmatics : –. Giora, R., N. Balaban, O. Fein, and I. Alkabets (). Negation as positivity in disguise. In H. L. Colston and A. Katz (eds.), Figurative Language Comprehension: Social and Cultural Influences, –. Hillsdale, NJ: Erlbaum, Giora, R., O. Fein, K. Aschkenazi, and I. Alkabets-Zlozover (). Negation in context: A functional approach to suppression. Discourse Processes : –. Giora, R., V. Heruti, N. Metuki, and O. Fein (). “When we say no we mean no”: Interpreting negation in vision and language. Journal of Pragmatics : –. Givón, Talmy (). Negation in language: Pragmatics, function, ontology. In Peter Cole (ed.), Syntax and Semantic, Vol : Pragmatics, –. New York: Academic Press.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Givón, Talmy (). On Understanding Grammar. New York: Academic Press. Givón, Talmy (). Syntax: A Functional-Typological Approach (Vols.  & ). Amsterdam: John Benjamins. Glenberg, A. M., and V. Gallese (). Action-based language: A theory of language acquisition, comprehension, and production. Cortex  (): –. Glenberg, A. M., and M. P. Kaschak (). Grounding language in action. Psychonomic Bulletin and Review : –. Glenberg, A. M., D. A. Robertson, J. L. Jansen, and M. C. Johnson-Glenberg (). Not propositions. Cognitive Systems Research : –. Godard, Danièle (). French negative dependency. In Francis Corblin and Henriëtte de Swart (eds.), Handbook of French Semantics, –. Stanford, CA: CSLI Publications. Goffman, Erving (). The Presentation of Self in Everyday Life. New York: Anchor Books. Gökgöz, Kadir (). Negation in Turkish Sign Language: The syntax of nonmanual markers. Sign Language & Linguistics  (): –. Goldin-Meadow, Susan (). The Resilience of Language: What Gesture Creation in Deaf Children Can Tell us about Language-Learning in General. New York: Psychology Press. González-Fuente, Santiago, Susagna Tubau, M.Teresa Espinal, and Pilar Prieto (). Is there a universal answering strategy for rejecting negative propositions? Typological evidence on the use of prosody and gesture. Frontiers in Psychology : . Goodhue, Daniel (). On Asking and Answering Biased Polar Questions. Ph.D. dissertation, McGill University. Goodhue, Daniel, and Michael Wagner (). Intonation, yes and no. Glossa: A Journal of General Linguistics  (): . Goossens, Louis (). Modal shifts and predication types. In J. van der Auwera and L. Goossens (eds.), Ins and Outs of the Predication, –. Dordrecht: Foris. Gordon, David, and George Lakoff (). Conversational postulates. In Peter Cole and Jerry L. Morgan (eds.), Syntax and Semantics : Speech Acts, –. New York: Academic Press. Goro, Takuya (). The Emergence of Universal Grammar in the Emergence of Language: The Acquisition of Japanese Logical Connectives and Positive Polarity. Ph.D. dissertation, University of Maryland. Goro, Takuya, and Sachie Akiba (). The acquisition of disjunction and positive polarity in Japanese. In V. Chand, A. Kelleher, A. J. Rodriguez, and B. Schmeiser (eds.), Proceedings of the rd West Coast Conference on Formal Linguistics, –. Somerville, MA: Cascadilla Press. Gough, P. B. (). Grammatical transformations and speed of understanding. Journal of Verbal Learning and Verbal Behavior : –. Gough, P. B. (). The verification of sentences: The effects of delay of evidence and sentence length. Journal of Verbal Learning and Verbal Behavior : –. Green, Lisa (). African American English: A Linguistic Introduction. Cambridge: Cambridge University Press. Green, Lisa (). Language and the African American Child. Cambridge: Cambridge University Press. Green, Lisa (). Force, focus, and negation in African American English. In Raffaella Zanuttini and Laurence R. Horn (eds.), Micro-Syntactic Variation in North American English, –. Oxford: Oxford University Press. Green, Mitchell (). Irony as expression (of a sense of the absurd). The Baltic International Yearbook of Cognition, Logic and Communication. Vol. : Irony. . Greenberg, Joseph H. (). Some universals of grammar with particular reference to the order of meaningful elements. In Joseph H. Greenberg (ed.), Universals of Language, –. Cambridge, MA: MIT Press.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Greene, Judith M. (). Syntactic form and semantic function. Quarterly Journal of Experimental Psychology : –. Gregerson, Kenneth (). Predicate and Argument in Rengao Grammar. Dallas: Summer Institute of Linguistics. Grice, Herbert Paul (). Logic and conversation. In Peter Cole and Jerry L. Morgan (eds.), Syntax and Semantics : Speech Acts, –. New York: Academic Press. Reprinted in Grice, Studies in the Way of Words, –. Grice, Herbert Paul (). Further notes on logic and conversation. In Peter Cole (ed.), Syntax and Semantics : Pragmatics, –. New York: Academic Press. Reprinted in Grice, Studies in the Way of Words, –. Grice, Herbert Paul (). Studies in the Way of Words. Cambridge, MA: Harvard University Press. Grinstead, John (). Subjects, Sentential Negation and Imperatives in Child Spanish and Catalan. Ph.D. dissertation, UCLA. Grodzinsky, Yosef (). Language deficits and the theory of syntax. Brain & Language : –. Grodzinsky, Yosef (). The neurology of syntax: Language use without Broca’s area. Behavioral and Brain Sciences  (): –. DOI: ./SX. Grodzinsky, Yosef, Isabelle Deschamps, Peter Pieperhoff, Francesca Ianilli, Galit Agmon, Yonatan Loewenstein, and Katrin Amunts (). Logical negation mapped onto the brain. Brain Struct Funct. Grodzinsky, Yosef, and Andrea Santi (). The battle for Broca’s region. Trends in Cognitive Sciences  (): –. Grodzinsky, Yosef, Galit Agmon, and Isabelle Deschamps (). Quantifier polarity processing and numerical comparison: fMRI results show striking modularity. Paper presented at the Society for Neurobiology of Language, London, August. Grodzinsky, Yosef, Galit Agmon, Kedem Snir, Isabelle Deschamps, and Yonatan Loewenstein (). The analysis of less-comparatives: Evidence from the processing cost of downward entailingness. Proceedings of Sinn und Bedeutung . Berlin/Potsdam. Groenendijk, Jeroen, and Martin Stokhof (). Studies on the Semantics of Questions and the Pragmatics of Answers. Ph.D. dissertation, University of Amsterdam. Grohmann, Kleanthes K. (). Top issues in questions: Topics-topicalization-topicalizability. In Lisa Lai-Shen Cheng and Norbert Corver (eds.), Wh-Movement: Moving On, –. Cambridge, MA, and London: MIT Press. Grüter, Theres, Moti Lieberman, and Andrea Gualmini () Acquiring the scope of disjunction and negation in L: A bidirectional study of learners of Japanese and English, Language Acquisition  (): –. Gualmini, Andrea (). Some knowledge children don’t lack. Linguistics : –. Guasti, Maria Teresa (). Verb syntax in Italian child grammar: Finite and nonfinite verbs. Language Acquisition : –. Guasti, Maria Teresa (). Acquisition of Italian interrogatives. In Harald Clahsen (ed.), Generative Perspectives on Language Acquisition, –. Amsterdam: John Benjamins. Guasti, Maria Teresa, Rosalind Thornton, and Kenneth Wexler (). Negation in children’s questions: The case of English. In D. MacLaughlin and S. McEwen (eds.), Proceedings of the thAnnual Boston University Conference on Language Development, –. Somerville, MA: Cascadilla Press. Guerzoni, Elena (). Why Even Ask? On the Pragmatics of Questions and the Semantics of Answers. Ph.D. dissertation, MIT. Guerzoni, Elena (). Even-NPIs in yes/no questions. Natural Language Semantics  (): –. Guidetti, Michèle (). Pragmatic study of refusal and agreement messages in young French children. Journal of Pragmatics  (): –. Guidetti, Michèle (). The emergence of pragmatics: Forms and functions of conventional gestures in young French children. First Language  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Guidetti, Michèle (). Yes or no? How do young children combine gestures and words to agree and refuse. Journal of Child Language : –. Guidetti, Michèle, and Aliyah Morgenstern (). The gesture–sign interface in language acquisition. Language, Interaction, Acquisition  (): –. Guillaume, Antoine (). Negation in Tacana (Amazonian Bolivia): Descriptive, typological and historical perspectives. Handout Syntax of the World’s Languages VII, Mexico City. Güldemann, Tom (). The genesis of verbal negation in Bantu and its dependency on functional features and clause types. In Jean-Marie Hombert and Larry Hyman (eds.), Bantu Historical Linguistics: Theoretical and Empirical Linguistics, –. Stanford, CA: Centre for the Study of Language and Information. Güldemann, Tom (). The Macro-Sudan belt: Towards identifying a linguistic area in northern sub-Saharan Africa. In Bernd Heine and Derek Nurse (eds.). A Linguistic Geography of Africa, –. Cambridge: Cambridge University Press. Gutiérrez-Rexach, Javier (). The semantic basis of NPI licensing in questions. In Benjamin Bruening (ed.), Proceedings of the th Student Conference in Linguistics, –. MIT Working Papers in Linguistics . Cambridge, MA: MIT. Gutiérrez-Rexach, Javier, and Patricia Andueza (). Degree restriction in Spanish exclamatives. In Luis A. Ortiz-López (ed.), Selected Proceedings of the th Hispanic Linguistics Symposium, –. Sommerville, MA: Cascadilla Press. Gutzmann, Daniel, and Elena Castroviejo-Miró (). The dimensions of Verum. In Olivier Bonami and Patricia Cabredo Hofherr (eds.), Empirical Issues in Syntax and Semantics , –. . Guy, G. (). Explanation in variable phonology: An exponential model of morphological constraints. Language Variation and Change  (): –. Gyuris, Beáta (). New perspectives on bias in polar questions: A study of Hungarian -e, International Review of Pragmatics  (): –. Hackl, Martin (). Comparative quantifiers. Ph.D. dissertation, MIT. Haddican, William (). Sentence polarity and word order in Basque. The Linguistic Review : –. Haeberli, Eric (). Adjuncts and the syntax of subjects in Old and Middle English. In Susan Pintzuk, George Tsoulas, and Anthony Warner (eds.), Diachronic Syntax: Models and Mechanisms, –. Oxford: Oxford University Press. Haegeman, Liliane (). The Syntax of Negation (Cambridge Studies in Linguistics ). Cambridge: Cambridge University Press. Haegeman, Liliane (). Negative preposing, negative inversion, and the split CP. In Laurence R. Horn and Yasuhiko Kato (eds.), Negation and Polarity: Syntactic and Semantic Perspectives, –. Oxford: Oxford University Press. Haegeman, Liliane (). Conditionals, factives and the left periphery. Lingua : –. Haegeman, Liliane (). West Flemish verb-based discourse markers and the articulation of the speech act layer. Studia Linguistica  (): –. Haegeman, Liliane, and Anne Breitbarth (). “En” en is níet wat we dachten: A Flemish discourse particle. MIT Working Papers in Linguistics : –. Haegeman, Liliane, and Virginia Hill (). The syntactization of discourse. In Raffaella Folli, Christina Sevdali, and Robert Truswell (eds.), Syntax and its Limits, –. Oxford: Oxford University Press. Haegeman, Liliane, and Terje Lohndal (). Negative Concord and (Multiple) Agree: A case study of West Flemish. Linguistic Inquiry  (): –. Haegeman, Liliane, and Raffaella Zanuttini (). Negative heads and the Neg criterion. The Linguistic Review : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Haegeman, Liliane, and Raffaella Zanuttini (). Negative concord in West Flemish. In Adriana Belletti and Luigi Rizzi (eds.), Parameters and Functional Heads: Essays in Comparative Syntax, –. Oxford: Oxford University Press. Hagège, Claude (). La structure des langues. Paris: Presses Universitaires de France. Hagemeijer, Tjerk (). Double-headed negation in Santome. Journal of Portuguese Linguistics  (): –. Hagiwara, Hiroko (). The breakdown of functional categories and the economy of derivation. Brain and Language  (): –. Hagoort, Peter, Colin Brown, and Jolanda Groothusen (). The syntactic positive shift as an ERP measure of syntactic processing. Lang. Cogn. Process : –. Hagoort, Peter, Lea Hald, Marcel Bastiaansen, and Karl Magnus Petersson (). Integration of word meaning and world knowledge in language comprehension. Science : –. Hagstrom, Paul (). Decomposing questions. Ph.D. dissertation, Massachusetts Institute of Technology, Cambridge, MA. Hagstrom, Paul (). Implications of child errors for the syntax of negation in Korean. Journal of East Asian Linguistics : –. Haida, Andreas (). The Indefiniteness and Focusing of Wh-Words. Ph.D. dissertation, HumboldtUniversität zu Berlin, Berlin. Haida, Andreas, and Sophie Repp (). The intervention effect: Focus alternatives or indefinite alternatives? Experimental evidence. In Maria Aloni, Michael Franke, and Floris Roelofsen (eds.), Proceedings of the th Amsterdam Colloquium, –. Amsterdam: ILLC. Hajičová, Eva (). Topic-focus articulation. A matter of langue or parole! The case of negation. In R. Sackmann (ed.), Papers in Honour of Hans-Heinrich Lieb on Occasion of his th Birthday, –. Amsterdam: John Benjamins. Hald, L., M. Kutas, T. P. Urbach, and B. Pahrhizkari (). The N is not a brainwave: Negation and the N effects for true and false sentences. Poster presented at the annual meeting of the Cognitive Neuroscience Society, New York. Hale, Mark (). Diachronic syntax. Syntax : –. Hamann, Cornelia (). Negation, infinitives and heads. In S. Powers and C. Hamann (eds.), The Acquisition of Scrambling and Cliticization, –. Dordrecht: Kluwer Academic Publishers. Hamari, Arja (). Negation in the history of the Mordvin languages. In David Willis, Christopher Lucas, and Anne Breitbarth (eds.), The History of Negation in the Languages of Europe and the Mediterranean, Vol. : Case Studies, –. Oxford: Oxford University Press. Hamawand, Zeki (). The Semantics of English Negative Prefixes. London: Equinox. Hamblin, Charles (). Questions in Montague English. Foundations of Language : –. Hampton, Robert R., Aaron Zivin, and Elisabeth A. Murray (). Rhesus monkeys (Macaca mulatta) discriminate between knowing and not knowing and collect information as needed before acting. Animal Cognition : –. Han, Chung-Hye (). Interpreting interrogatives and rhetorical questions. Lingua : –. Han, Chung-Hye, and Laura Siegel (). Syntactic and semantic conditions of NPI licensing in questions. In Brian Agbayani and Sze-Wing Tang (eds.), Proceedings of the West Coast Conference in Formal Linguistics , –. Stanford, CA: CSLI Publications,. Han, Chung-Hye, Jeffery Lidz, and Julien Musolino (). V-raising and grammar competition in Korean: Evidence from negation and quantifier scope. Linguistic Inquiry  (): –. Han, H., and M.-K. Park. (). The syntax of negation in Korean and its development in child language. In H. Hoji (ed.) ESCOL, –. Stanford, CA: CSLI. Hanakawa, Takashi, Manabu Honda, Nobukatsu Sawamoto, Tomohisa Okada, Yoshiharu Yonekura, Hidena Fukuyama, and Hiroshi Shibasaki (). The role of rostral Brodmann area  in mentaloperation tasks: An integrative neuroimaging approach. Cerebral Cortex  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Hankamer, Jorge (). Deletion in Coordinate Structures. New York: Garland. Hansen, Maj-Britt Mosegaard (). The grammaticalization of negative reinforcers in Old and Middle French: A discourse-functional approach. In Maj-Britt Mosegaard Hansen and Jacqueline Visconti (eds.), Current Trends in Diachronic Semantics and Pragmatics, –. Bingley: Emerald. Hansen, Maj-Britt Mosegaard (). On the evolution of temporal n-words in Medieval French. Language Sciences : –. Hansen, Maj–Britt Mosegaard (). Negation in the history of French. In David Willis, Christopher Lucas, and Anne Breitbarth (eds.), The History of Negation in the Languages of Europe and the Mediterranean. Vol. : Case Studies, –. Oxford: Oxford University Press. Hansen, Maj-Britt Mosegaard (). The grammaticalization of negative indefinites: The case of the temporal-aspectual n-words ‘mais’ and ‘plus’ in Medieval French. In Maj-Britt Mosegaard Hansen and Jacqueline Visconti (eds.), Diachrony of Negation, –. Amsterdam: Benjamins. Hansen, Maj-Britt Mosegaard (). The expression of clause negation: From Latin to Early French. In Anne Carlier and Céline Guillot-Barbance (eds.), Latin: Tardif/français ancien: Continuïtés et ruptures, –. Berlin: de Gruyter Mouton. Hansen, Maj-Britt Mosegaard, and Jacqueline Visconti (). On the diachrony of “reinforced” negation in French and Italian. In Corrine Rossari, Claudia Ricci, and Adriana Spiridon (eds.), Grammaticalisation and Pragmatics: Facts, Approaches, Theoretical Issues, –. Bingley: Emerald Publishing. Hansen, Maj-Britt Mosegaard, and Jacqueline Visconti (eds.) (). The Diachrony of Negation. Amsterdam: Benjamins. Hardt, Daniel (). Verb Phrase Ellipsis: Form, Meaning, and Processing. University of Pennsylvania, Ph.D. dissertation. Hare, Brian, Josep Call, and Michael Tomasello (). Do chimpanzees know what conspecifics know? Animal Behaviour : –. Hare, Brian, Josep Call, Bryan Agnetta, and Michael Tomasello (). Chimpanzees know what conspecifics do and do not see. Animal Behaviour : –. Harper J., S. F. Malone, and E. M. Bernat (). Theta and delta band activity explain N and P ERP component in go/NoGo task. Clin Neurophysiol : –. Harrell, Richard (). A Short Reference Grammar of Moroccan Arabic. Washington, DC: Georgetown University Press. Harris, Tony, and Ken Wexler (). The optional-infinitive stage in child English: Evidence from negation. In H. Clahsen (ed.), Generative Perspectives on Language Acquisition, –. Amsterdam: John Benjamins. Harrison, Simon (). Grammar, Gesture, and Cognition: The Case of Negation in English. Université Michel de Montaigne Bordeaux . Harrison, Simon (). Evidence for node and scope of negation in coverbal gesture. Gesture  (): –. Harrison, Simon (). The Impulse to Gesture: Where Language, Minds, and Bodies Intersect. Cambridge: Cambridge University Press. Hartung, Simone (). Forms of Negation in Polar Questions. MA thesis, University of Tübingen. Hartzler, Margaret (). Sentani. In Kahrel and van den Berg (eds.), Typological Studies in Negation, –. Haspelmath, Martin (). Indefinite Pronouns. Oxford: Oxford University Press. Haspelmath, Martin (). The relevance of extravagance: A reply to Bart Geurts. Linguistics : –. Haspelmath, Martin (). Coordination. In Timothy Shopen (ed.), Language Typology and Syntactic Description, –. Cambridge: Cambridge University Press. Haspelmath, Martin (). Frequency vs. iconicity in explaining grammatical asymmetries. Cognitive Linguistics  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Hasson, Uri, and Sam Glucksberg (). Does understanding negation entail affirmation? An examination of negated metaphors. Journal of Pragmatics  (): –. Hauser, Eric (). Stability and change in one adult’s second language English negation. Language Learning  (): –. Heath, Jeffrey (). Basic Materials in Mara: Grammar, Texts and Dictionary. Canberra: Research School of Pacific Studies, Australian National University. Heath, Jeffrey (). A Grammar of Tamashek (Tuareg of Mali). Berlin: Mouton de Gruyter. Hedberg, Nanzy, and Juan Sosa (). Pitch contours in negative sentences. th International Congress of Phonetic Sciences. Barcelona. Heim, Irene (). The Semantics of Definite and Indefinite Noun Phrases. Ph.D. dissertation, University of Massachusetts, Amherst. Heim, Irene (). On the projection problem for presuppositions. In P. H. Portner (ed.), Formal Semantics: The Essential Readings, –. Oxford: Blackwell. Heim, Irene (). A note on negative polarity and downward entailingness. In Charles Jones and Peter Sells (eds.), Proceedings of the North East Linguistics Society , –. Amherst, MA: GLSA. Heim, Irene (). Presupposition projection and the semantics of attitude verbs. Journal of Semantics  (): –. Heim, Irene (). Degree operators and scope. In Brendan Jackson and Tanya Matthews (eds.), Proceedings of SALT X, –. Cornell University, Ithaca, NY: CLC Publications. Heim, Irene (). Degree operators and scope. In Caroline Féry and Wolfgang Sternefeld (eds.), Audiatur Vox Sapientiae: A Festschrift for Arnim von Stechow, –. Studia Grammatica . Berlin: Akademie Verlag. Heim, Irene (). Little. In Masayuki Gibson and Jonathan Howell (eds.), Proceedings of Semantics and Linguistic Theory : –. Heim, Irene, and Angelika Kratzer (). Semantics in Generative Grammar. Malden, MA: Blackwell. Heim, Stefan, Katrin Amunts, Dan Drai, Simon Eickhoff, Sara Hautvast, and Yosef Grodzinsky (). The language–number interface in the brain: A complex parametric study of quantifiers and quantities. Frontiers in Evolutionary Neuroscience  (): –. Hendriks, Bernadet (). Negation in Jordanian Sign Language: A cross-linguistic perspective. In Pamela Perniss, Roland Pfau, Markus Steinbach (eds.), Visible Variation: Cross-linguistic Studies in Sign Language Structure, –. Berlin: Mouton de Gruyter. Henry, Alison (). Belfast English and Standard English: Dialect Variation and Parameter Setting. New York: Oxford University Press. Henry, Alison (). Acquiring language from variable input: Subject–verb agreement and negative concord in Belfast English. Linguistic Variation  (): –. Henry, Desmond (). The Logic of Saint Anselm. Oxford: Clarendon Press. Herbert, C., and A. Kübler (). Dogs cannot bark: Event-related brain responses to true and false negated statements as indicators of higher-order conscious processing. PloS One , e. Herburger, Elena (). What Counts: Focus and Quantification. Cambridge, MA: MIT Press. Herburger, Elena (). The negative concord puzzle revisited. Natural Language Semantics : –. Herburger, Elena (). A note of Spanish ni siquiera, even, and the analysis of NPIs. Probus : –. Hernanz, M. Lluïsa (). Emphatic polarity and C in Spanish. In Laura Bruges (ed.), Studies in Spanish Syntax, –. Venice: Cafoscarina. Herrmann, Esther, and Josep Call (). Are there geniuses among the apes? Phil. Trans. R. Soc. B : –. Heyes, Cecilia (). Animal mindreading: What’s the problem? Psychonomic Bulletin & Review : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Heyting, Arend (). Intuitionism: An Introduction. Amsterdam: North-Holland Publishing. Hicks, Glyn (). The Derivation of Anaphoric Relations. Amsterdam: John Benjamins. Higginbotham, James (). The logic of perceptual reports: An extensional alternative to Situation Semantics. Journal of Philosophy : –. Higginbotham, James (). Is semantics necessary? Proceedings of the Aristotelian Society : –. Higginbotham, James (). Mass and count quantifiers. Linguistics & Philosophy : –. Higginbotham, James (). Perceptual reports revisited. In Kumiko Murasagi and Robert Stainton (eds.), Philosophy and Linguistics, –. Boulder, CO: Westview Press. Higginbotham, James (). On events in linguistic semantics. In James Higginbotham, Fabio Pianesi, and Achille C. Varzi (eds.), Speaking of Events, –. New York: Oxford University Press. Hill, Andrew, Emma Collier-Baker, and Thomas Suddendorf (). Inferential reasoning by exclusion in great apes, lesser apes, and spider monkeys. Journal of Comparative Psychology : –. Hill, Virginia (). Vocatives: How Syntax Meets with Pragmatics. Leiden: Brill. Hintikka, Jaakko (). Knowledge and Belief. Ithaca, NY: Cornell University Press. Hintikka, Jaakko (). Epistemic logic and the methods of philosophical analysis. Australasian Journal of Philosophy : –. Hintikka, Jaako (). Semantics for propositional attitudes. In J. W. Davis, D. J. Hockney, and W. K. Wilson (eds.), Philosophical Logic, –. Dordrecht: Reidel. Hiraiwa, Ken (). Multiple agreement and the defective intervention effect. In O. Matsushansky (ed.), The Proceedings of the MIT-Harvard Joint Conference (HUMIT ), –. Cambridge, MA: MITWPL. Hiramatsu, Kazuko (). Children’s judgments of negative questions. Language Acquisition  (): –. Hirata, Satoshi, and Tetsuro Matsuzawa (). Tactics to obtain a hidden food item in chimpanzee pairs (Pan troglodytes). Animal Cognition : –. Hirschberg, Julia, and Cinzia Avesani (). Prosodic disambiguation in English and Italian. In Antonis Botinis (ed.), Intonation: Analysis, Modelling and Technology, –. Dordrecht: Kluwer. Hitzeman, Janet (). Aspect and adverbials. Proceedings of Semantics and Linguistic Theory : –. Hochmann, J.-R. (). Word frequency, function words and the second Gavagai problem. Cognition : –. Hochmann, J.-R., and J. M. Toro (submitted). Pre-verbal negation and abstract relations in infancy. Hochmann, J.-R., S. Carey, and J. Mehler (). Infants learn a rule predicated on the relation same but fail to simultaneously learn a rule predicated on the relation different. Cognition : –. Hochmann, J.-R., A. D. Endress and J. Mehler (). Word frequency as a cue to identify function words in infancy. Cognition : –. Hochmann, J.-R., S. Mody, and S. Carey (). Infants’ representations of same and different in match- and non-match-to sample. Cognitive Psychology : –. Hochmann, J.-R., S. Benavides-Varela, M. Nespor, and J. Mehler (). Vowels and consonants in early language acquisition. Developmental Science : –. Hochmann, J.-R., S. Benavides-Varela, A. Fló, M. Nespor, and J. Mehler (). Bias for vocalic over consonantal information in -month-olds. Infancy : –. Hoeksema, Jack (). On the grammaticalization of negative polarity items. In Proceedings of Berkeley Linguistic Society, –. Berkeley. Hoeksema, Jack (a). Aantekeningen bij ooit, deel : De opkomst van niet polair ooit. Tabu  (): –. Hoeksema, Jack (b). Blocking effects and polarity sensitivity. In Jelle Gerbrandy, Maarten Marx, Maarten de Rijke, and Yde Venema (eds.), JFAK: Essays Dedicated to Johan van Benthem on the Occasion of his th Birthday. Amsterdam: Vossiuspers. Hoeksema, Jack (). Negative polarity items: Triggering, scope, and C-command. In L. Horn and Y. Kato (eds.), Negation and Polarity, –. Oxford: Oxford University Press.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Hoeksema, Jack (). Minimaliseerders in het standaard-Nederlands [Minimizers in standard Dutch]. Tabu  (/): –. Hoeksema, Jack (). Jespersen recycled. In Elly van Gelderen (ed.), Cyclical Change, –. Amsterdam/Philadelphia: John Benjamins. Hoeksema, Jack (). Dutch ENIG: From nonveridicality to downward entailment. Natural Language and Linguistic Theory  (): –. Hoeksema, Jack (). On the natural history of negative polarity items. Linguistic Analysis  (–): –. Hoeksema, Jack (). Neg-raising and log-distance licensing of negative polarity items. In D. Ziegeler and B. Zhiming (eds.), Negation and Contact, –. Amsterdam: Benjamins. Hoeksema, Jack, and Hotze Rullmann (). Scalarity and polarity: A study of scalar adverbs as polarity items. In Jack Hoeksema, Hotze Rullmann, Víctor Sánchez-Valencia, and Ton van der Wouden (eds.), Perspectives on Negation and Polarity Items, –. Amsterdam: John Benjamins. Hoepelman, J., and C. Rohrer (). Remarks on noch and schon in German. In P. J. Tedeschi and A. Zaenen (eds.), Syntax and Semantics : Tense and Aspect, pp. –. New York: Academic Press. Hoffmann, Maria (). Negatio Contrarii: A Study of Latin Litotes. Assen: Van Gorcum. Hohepa, Patrick W. (). Not in English and kore and eehara in Maori. Te Reo : –. Höhle, Tilman (). Über Verum Fockus im Deutschen. Linguistische Berichte Sonderhefte : –. Hoji, Hajime (). Logical Form Constraints and Configurational Structures in Japanese. Ph.D. dissertation, University of Washington. Holmberg, Anders (). On the syntax of yes and no in English. Ms., Newcastle University. Holmberg, Anders (). The syntax of answers to polar questions in English and Swedish. Lingua : –. Holmberg, Anders (). The Syntax of Yes and No. Oxford: Oxford University Press. Homer, Vincent (). Epistemic modals: High ma non troppo. In Seda Kan, Claire MooreCantwell, and Robert Staubs (eds.), Proceedings of the th Annual Meeting of the North Eastern Linguistic Society, Vol. , –. Amherst, MA: GLSA. Homer, Vincent (). Neg raising and positive polarity: The view from modals. Semantics and Pragmatics  (): –. Hömke, Lars, Katrin Amunts, L. Bönig, C. Fretz, Ferdinand Binkofski, Karl Zilles, and B. Weder (). Analysis of lesions in patients with unilateral tactile agnosia using cytoarchitectonic probabilistic maps. Human Brain Mapping  (): –. Honcoop, Martin (). Dynamic Excursions on Weak Islands. Ph.D. dissertation, University of Leiden, Leiden. Honti, László (). Die Negation im Uralischen. Linguistica Uralica : –, –, –. Hooper, Joan, and Sandra Thompson (). On the applicability of root transformations. Linguistic Inquiry  (): –. Hopkins, Jasper (). A Companion to the Study of St. Anselm. Minneapolis: University of Minnesota Press. Hopper, Paul J., and Elisabeth Closs Traugott (). Grammaticalization. nd edn. Cambridge: Cambridge University Press. Hörmann, H. (). Semantic factors in Negation. Psychologische Forschung : –. Horn, Laurence R. (). A presuppositional analysis of only and even. In R. Binnick et al. (eds.), Papers from the th Annual Meeting of the Chicago Linguistics Society, –. Chicago: Chicago Linguistics Society. Horn, Laurence R. (). Negative transportation: Unsafe at any speed? CLS : –. Horn, Laurence R. (). On the Semantic Properties of the Logical Operators in English. Ph.D. dissertation, UCLA. Horn, Laurence R. (). Neg-raising predicates: Toward an explanation. CLS : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Horn, Laurence R. (). On the Semantic Properties of Logical Operators in English. Bloomington: IULC. Horn, Laurence R. (). Remarks on Neg-Raising. In Peter Cole (ed.), Syntax and Semantics : Pragmatics, –. New York: Academic Press. Horn, Laurence R. (). Toward a new taxonomy for pragmatic inference: Q-based and R-based implicature. In Deborah Schiffrin (ed.), Meaning, Form and Use in Context: Linguistic Applications (GURT’), –. Washington, DC: Georgetown University Press. Horn, Laurence R. (). Metalinguistic negation and pragmatic ambiguity. Language : –. Horn, Laurence R. (). A Natural History of Negation. Chicago: University of Chicago Press (reprint  Stanford, CA: CSLI). Horn, Laurence R. (). Review: Showdown at truth-value gap: Burton-Roberts on presupposition (reviewed work: The Limits to Debate: A Revised Theory of Semantic Presupposition by Noel Burton-Roberts). Journal of Linguistics  (): –. Horn, Laurence R. (). Duplex negatio affirmat: The economy of double negation. In Lise M. Dobrin, Lynn Nichols, and Rose M. Rodriguez (eds.), Papers from the th Regional Meeting of the Chicago Linguistic Society. Part Two: The Parasession on Negation, –. Chicago: Chicago Linguistic Society. Horn, Laurence R. (). The said and the unsaid. In Chris Baker and David Dowty (eds.), SALT II: Proceedings from the Second Conference on Semantics and Linguistic Theory. The Ohio State University Working Papers in Linguistics : –. Horn, Laurence R. (). Exclusive company: Only and the dynamics of vertical inference. Journal of Semantics : –. Horn, Laurence R. (a). A Natural History of Negation. Stanford, CA: CSLI. Horn, Laurence R. (b). The logic of logical double negation. In Yasuhiko Kato (ed.), Proceedings of the Sophia Symposium on Negation, –. Tokyo: Sophia University. Horn, Laurence R. (c). Flaubert triggers, squatitive negation, and other quirks of grammar. In Jack Hoeksema, Hotze Rullmann, Víctor Sánchez-Valencia, and Ton van der Wouden (eds.), Perspectives on Negation and Polarity Items, –. Amsterdam: John Benjamins. Horn, Laurence R. (). Uncovering the un-word: A study in lexical pragmatics. Sophia Linguistica : –. Horn, Laurence R. (). Implicature. In Laurence R. Horn and Gregory Ward (eds.), The Handbook of Pragmatics, –. Oxford: Blackwell. Horn, Laurence R. (a). Airport ‘ revisited: Toward a unified indefinite any. In Gregory N. Carlson and Francis Jeffry Pelletier (eds.), Reference and Quantification: The Partee Effect, –. Stanford, CA: CSLI. Horn, Laurence R. (b). An un-paper for the unsyntactician. In S. Mufwene, E. Francis, and R. Wheeler (eds.), Polymorphous Linguistics: Jim McCawley’s Legacy, –. Cambridge, MA: MIT Press. Horn, Laurence R. (a). Multiple negation in English and other languages. In Laurence R. Horn (ed.), The Expression of Negation, –. Berlin and New York: Mouton de Gruyter. Horn, Laurence R. (ed.) (b). The Expression of Negation. Berlin: Mouton de Gruyter. Horn, Laurence R. (). Histoire d’*O: Lexical pragmatics and the geometry of opposition. In J.-Y. Béziau and G. Payette (eds.), The Square of Opposition: A General Framework for Cognition, –. Bern: Peter Lang. Horn, Laurence R. (). Foreword Classical NEG Raising: The first  years. In C. Collins and P. Postal (eds.), Classical NEG Raising: An Essay on the Syntax of Negation, ix–xvii. Cambridge, MA: MIT Press. Horn, Laurence R. (a). (Neo-)Classical Neg-raising and strict licensing: What’s at issue. Paper presented at the Amsterdam Colloquium. Horn, Laurence R. (b). On the contrary: Disjunctive syllogism and pragmatic strengthening. In A. Koslow and A. Buchsbaum (eds.), The Road to Universal Logic, –. Heidelberg: Birkhäuser (Springer).

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Horn, Laurence R. (). Licensing NPIs: Some negative (and positive) results. In P. Larrivée and C. Lee (eds.), Negation and Polarity: Experimental Perspectives, –. Berlin: Springer. Horn, Laurence R. (a). Lie-toe-tease: Double negatives and unexcluded middles. Philosophical Studies : –. Horn, Laurence R. (b). The singular square: Contrariety and double negation from Aristotle to Homer. In Joanna Blochowiak, Cristina Grisot, Stephanie Durrlemann, and Christopher Laenzlinger (eds.), Formal Models in the Study of Language, –. Cham: Springer. Horn, Laurence R. () Contradiction. In E. Zalta (ed.), Stanford Encyclopedia of Philosophy, . Horn, Laurence R., and Samuel Bayer (). Short-circuited implicature: A negative contribution. Linguistics and Philosophy : –. Horn, Laurence R., and Young-Suk Lee (). How many any’s diagnosing the diagnostics. Paper presented at the annual meeting of the LSA. Horn, Laurence R., and Heinrich Wansig (). Negation. In E. Zalta, U. Nodelman, C. Allen, and R. L. Anderson (eds.), Stanford Encyclopedia of Philosophy. Digital book: . Horrocks, Geoffrey (). Ouk ísmen oudén: Negative Concord and negative polarity in the history of Greek. Journal of Greek Linguistics  (): –. Hovdhaugen, Even, and Ulrike Mosel (eds.) (). Negation in Oceanic Languages. Munich: LINCOM EUROPA. Hoyt, Frederic (). Prosodic constituency and locality. In Levantine Arabic long-distance negative concord. In Reem Khamis-Dakwar and Karen Froud (eds.), Perspectives on Arabic Linguistics XXVI: Papers from the Annual Symposium on Arabic Linguistics (New York, ), –. Amsterdam: John Benjamins. Huang, Aijun, and Stephen Crain (). Polarity sensitive expressions in child Mandarin. Language Acquisition : –. Huddleston, Rodney, and Geoffrey K. Pullum (). The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Huddlestone, Kathleen M. (). Negative indefinites in Afrikaans. Ph.D. dissertation, Utrecht University. Huette, S., and S. Anderson (). Negation without symbols: The importance of recurrence and context in linguistic negation. Journal of Integrative Neuroscience : . Huettel, Scott A., Allen W. Song, and Gregory McCarthy (). Functional Magnetic Resonance Imaging. nd edn. Sunderland, MA: Sinauer Associates. Hulk, Aafke, and Natascha Müller (). Crosslinguistic influence at the interface between syntax and pragmatics. Bilingualism: Language and Cognition  (): –. Humberstone, Lloyd (). The revival of rejective negation. Journal of Philosophical Logic : –. Huster, R. J., S. Enriquez-Geppert, C. F. Lavallee, M. Falkenstein, C. S. Herrmann, et al. (). Electroencephalography of response inhibition tasks: Functional networks and cognitive contributions. Int J. Psychophysiol : –. Huttar, George, and Mary Huttar (). Ndyuka. London: Routledge. Huybregts, Rini (). De biologische kern van taal. In R. Huybregts and L. des Tombe (eds.), Verkenningen in Taal, –. Utrecht: Instituut A. W. de Groot. Hwang, Ju Hyeon (). Negative Sensitive Items: Conditions, Constraints, and Consequences. Ph.D. dissertation, University of Illinois. Hyman, Larry M. (). On the change from SOV to SVO: Evidence from Niger-Congo. In Charles N. Li (ed.), Word Order and Word Order Change, –. Austin: University of Texas Press. Iatridou, Sabine, and Ivy Sichel (). Negative DPs, A-movement, and scope diminishment. Linguistic Inquiry  (): –. Iatridou, Sabine, and Hedde Zeijlstra (). On the scopal interaction of negation and deontic modals. In Maria Aloni, Harald Bastiaanse, Tikitu de Jager, and Katrin Schulz (eds.), Logic, Language and Meaning, –. Amsterdam: Springer.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Iatridou, Sabine, and Hedde Zeijlstra (). Negation, polarity, and deontic modals. Linguistic Inquiry : –. Ibañez, Roberto (). Negation im Spanischen. Munich: Wilhelm Fink Verlag. Incurvati, Luca, and Julian J. Schlöder (). Weak rejection. Australasian Journal of Philosophy  (): –. Ingham, Richard (). Grammar change in Anglo-Norman and Continental French: The replacement of non-affirmative indefinite nul by aucun. Diachronica  (): –. Ingham, Richard (). Negation in the history of English. In Willis, Lucas, and Breitbarth (eds.), The History of Negation in the Languages of Europe and the Mediterranean, –. Inghilleri, M., A. Berardelli, G. Cruccu, and M. Manfredi (). Silent period evoked by transcranial stimulation of the human cortex and cervicomedullary junction. Journal of Physiology : –. Iordăchioaia, Gianina (). Negative Concord with Negative Quantifiers: A Polyadic Quantifier Approach to Romanian Negative Concord. University of Tubingen. Israel, Michael (). Negative polarity and phantom reference. In Berkeley Linguistic Society, –. Berkeley. Israel, Michael (). Polarity sensitivity as lexical semantics. Linguistics and Philosophy : –. Israel, Michael (). The Rhetoric of Grammar: Scalar Reasoning and Polarity Sensitivity. Ph.D. thesis, University of California San Diego. Israel, Michael (). Minimizers, maximizers and the rhetoric of scalar reasoning. Journal of Semantics  (): –. Israel, Michael (a). Polarity and Scalar Reasoning: The Pragmatic Basis of Grammar. Cambridge: Cambridge University Press. Israel, Michael (b). The Grammar of Polarity: Pragmatics, Sensitivity, and the Logic of Scales. Cambridge: Cambridge University Press. Jaakkola, Kelly (). Do animals understand invisible displacement? A critical review. Journal of Comparative Psychology : –. Jablonska, Patrycja (). Quirky n-words in Polish: NPIs, Negative Quantifiers or neither? In A. Dahl, K. Bentzen, and P. Svenonius (eds.), Proceedings of the th Scandinavian Conference of Linguistics  (). Jackendoff, Ray (). An interpretive theory of negation. Foundations of Language : –. Jackendoff, Ray (). On some questionable arguments about quantifiers and negation. Language : –. Jackendoff, Ray (). Semantic Interpretation in Generative Grammar. Cambridge, MA: MIT Press. Jackendoff, Ray (). Foundations of Language: Brain, Meaning, Grammar, Evolution. Oxford: Oxford University Press. Jacobs, Joachim (). Lexical decomposition in Montague grammar. Theoretical Linguistics : –. Jacobs, Joachim (). Negation. In A. von Stechow and D. Wunderlich (eds.), Semantics: An International Handbook of Contemporary Research, –. Berlin: Mouton de Gruyter. Jacobson, Pauline (). The short answer: Implications for direct compositionality (and vice versa). Language  (): –. Jacobson, Pauline (). Some people think there is neg raising, and some don’t: Neg raising meets ellipsis. Linguistic Inquiry : –. Jäger, Agnes (). History of German Negation. Amsterdam: John Benjamins. Jäger, Agnes (). Anything is nothing is something: On the diachrony of polarity types of indefinites. Natural Language and Linguistic Theory  (): –. Janssen, Theo (). Compositionality. (With an appendix by B. Partee.) In Johan Van Benthem and Alice Ter Meulen (eds.), Handbook for Logic and Language, –. Amsterdam: Elsevier; Cambridge, MA: MIT Press. Jaspers, Dany (). Operators in the Lexicon: On the Negative Logic of Natural Language. (Universiteit Leiden dissertation.) Utrecht: LOT.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Jayaseelan, K. A. (). Incomplete VP deletion and gapping. Linguistic Analysis : –. Jayaseelan, K. A. (). IP-internal topic and focus phrases. Studia Linguistica : –. Jayez, J. and L. M. Tovena (). Free-Choiceness and Non Individuation. Linguistics and Philosophy : –. Jelbert, Sarah A., Alex H. Taylor, and Russell D. Gray (). Reasoning by exclusion in New Caledonian crows (Corvus moneduloides) cannot be explained by avoidance of empty containers. Journal of Comparative Psychology : –. Jenner, Henry (). A Handbook of the Cornish Language. n.p.: David Nutt. [Project Gutenberg eBook: ] Jensen, B. (). Negative polarity items in English and Danish. In W. Abraham and J.-W. Zwart (eds.), Issues in Formal German(ic) Typology, –. Amsterdam: John Benjamins. Jeong, Sunwoo (). Intonation and sentence type conventions: Two types of rising declaratives. Journal of Semantics  (): –. Jespersen, Otto (–). A Modern English Grammar on Historical Principles ( vols.). Copenhagen: Munksgaard. Jespersen, Otto (). Negation in English and Other Languages. Copenhagen: A. F. Høst & Søn. Jespersen, Otto (). Language: Its Nature, Development and Origin. London: George Allen & Unwin. Jespersen, Otto (). The Philosophy of Grammar. London: Allen and Unwin. Johannessen, Janne Bondi (). Negative polarity verbs in Norwegian. Working Papers in Scandinavian Syntax : –. Johnson, I. R., B. M. Kopp, and R. E. Petty (). Just say no! (and mean it): Meaningful negation as a tool to modify automatic racial attitudes. Group Processes and Intergroup Relations : –. Johnson, Kyle (). When verb phrases go missing. In Lisa Cheng and Rint Sybesma (eds.), The First GLOT International State-of-the-Article Book, –. Berlin/New York: Mouton de Gruyter. Johnson, Kyle (). What VP ellipsis can do, and what it can’t, but not why. In Mark Baltin and Chris Collins (eds.), The Handbook of Contemporary Syntactic Theory, –. Oxford: Blackwell. Johnson-Laird, P. N. (). Mental Models: Towards a Cognitive Science of Language. Cambridge, MA: Harvard University Press. Johnston, Trevor (). A corpus-based study of the role of headshaking in negation in Auslan (Australian Sign Language): Implications for signed language typology. Linguistic Typology  (): –. Johnston, Trevor, and Adam Schembri (). Australian Sign Language: An Introduction to Sign Language Linguistics. Cambridge: Cambridge University Press. Jones, Bob Morris (). The Welsh Answering System. Berlin: Mouton de Gruyter. Jones, Sheila (a). Decoding a deceptive instruction. British Journal of Psychology : –. Jones, Sheila (b). The effect of a negative qualifier in an instruction. Journal of Verbal Learning and Verbal Behavior : –. Jones, Sheila (). Instructions, self-instructions and performance. The Quarterly Journal of Experimental Psychology : –. Jordens, Peter, and Christine Dimroth (). Finiteness in children and adults learning Dutch. In Natalia Gagarina and Insa Gülzow (eds.), The Acquisition of Verbs and their Grammar, –. Dordrecht: Springer. Joseph, John E., and Talbot J. Taylor (eds.) (). Ideologies of Language. London: Routledge. Joshi, Shrikant (). Affixal negation—direct, indirect and their subtypes. Syntaxe et sémantique : –. Jou, Jerwen (). The development of comprehension of double negation in Chinese children. Journal of Experimental Child Psychology : –. Just, Marcel A., and Patricia A. Carpenter (). Comprehension of negation with quantification. Journal of Verbal Learning and Verbal Behavior  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Just, Marcel A., and Patricia A. Carpenter (). Eye fixations and cognitive processes. Cognitive Psychology  (): –. Just, Marcel A., and H. H. Clark (). Drawing inferences from the presuppositions and implications of affirmative and negative sentences. Journal of Verbal Learning and Verbal Behavior : –. Kaan, E., and T. Swaab (a). Repair, revision and complexity in syntactic analysis: An electrophysiological differentiaton. J. Cogn. Neurosci. : –. Kaan, E., and T. Swaab (b). Electrophysiological evidence for serial sentence processing: A comparison between non-preferred and ungrammatical continuations. Cogn. Brain Res. : –. Kaan, E., A. Harris, E. Gibson, and P. Holcomb (). The P as an index of syntactic integration difficulty. Language and Cognitive Processes  (): –. Kadmon, Nirit, and Fred Landman (). Polarity sensitive any and free choice any. In Seventh Amsterdam Colloquium, –. Kadmon, Nirit, and Fred Landman (). Any. Linguistics and Philosophy  (): –. Kahneman, D. (). Thinking Fast and Slow. Toronto: Anchor Canada. Kahnemuyipour, Arsalan (). Negation is low in Persian: Evidence from nominalization. In Andrew A. Monti (ed.), Actes du congrès annuel de l’Association canadienne de linguistique , . Kahrel, Peter (). Aspects of Negation. Ph.D. dissertation, University of Amsterdam. Kahrel, Peter, and René van den Berg (eds.) (). Typological Studies in Negation. Amsterdam: Benjamins. Kaiser, Elsi (). Negation and the left periphery in Finnish. Lingua : –. Kallel, Amel, and Richard Ingham (). Evidence from a correspondence corpus for diachronic change in French indefinites –. In Maj-Britt Mosegaard Hansen and Jacqueline Visconti (eds.), The Diachrony of Negation, –. Amsterdam: Benjamins. Kallel, Amel, and Pierre Larrivée (under evaluation). The empirical reality of bridging contexts: A quantitative confirmation of strong polarity contexts as the transition between NPIs and n-words. Manuscript. Kamp, Hans (). A theory of truth and semantic representation. In J. Groenendijk, Th. Janssen, and M. Stokhof (eds.), Formal Methods in the Study of Language. Amsterdam: Mathematical Centre. Kamp, Hans, and Uwe Reyle (). From Discourse to Logic. Dordrecht: Kluwer Academic Publishers. Kandybowicz, Jason (). Varieties of Polarity Emphasis. Ms., University of Texas at Arlington. Karg, Katja, Martin Schmelz, Josep Call, and Michael Tomasello (). The goggles experiment: Can chimpanzees use self-experience to infer what a competitor can see? Animal Behaviour : –. Karimi, H., and F. Ferreira (). Good-enough linguistic representations and online cognitive equilibrium in language processing, The Quarterly Journal of Experimental Psychology  (): – Karttunen, Lauri (). Presuppositions of compound sentences. Linguistic Inquiry  (): –. Karttunen, Lauri (a). Presupposition and linguistic context. Theoretical Linguistics : –. Karttunen, Lauri (b). Until. In Tenth Regional Meeting of the Chicago Linguistic Society, –. Karttunen, Lauri (). Syntax and semantics of questions. Linguistics and Philosophy  (): –. Karttunen, Lauri, and Stanley Peters (). Requiem for presupposition. In Annual Meeting of the Berkeley Linguistics Society : –. Karttunen, Lauri, and Stanley Peters (). Conventional implicature. In Choon-Kyu Oh and David A. Dineen (eds.), Syntax and Semantics, Vol. : Presupposition, –. New York: Academic Press. Kato, Yasuhiko (). Negation in Classical Japanese. In Laurence R. Horn (ed.), The Expression of Negation, –. Berlin: de Gruyter. Katz, Jerrold (). Propositional Structure and Illocutionary Force. Cambridge, MA: MIT Press.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Katzir, Roni, and Raj Singh (). Constraints on the lexicalization of logical operators. Linguistics and Philosophy : –. Kaufman, J., G. Csibra, and M. H. Johnson (). Representing occluded objects in the human infant brain. Proceedings of the Royal Society of London. Series B: Biological Sciences  (suppl_): S–S. Kaufmann, Anita (). Negation and prosody in British English: A study based on the London– Lund Corpus. Journal of Pragmatics  (–): –. Kaup, Barbara (). The processing of negatives during discourse comprehension. In M. B. Shafto and P. Langley (eds.), Proceedings of the Nineteenth Conference of the Cognitive Science Society, –. Mahwah, NJ: Lawrence Erlbaum. Kaup, Barbara (). Negation and its impact on the accessibility of text information. Memory & Cognition : –. Kaup, Barbara (). How are pragmatic differences between affirmative and negative sentences captured in the processes and representations in language comprehension? In U. Sauerland and K. Yatsushiro (eds.), Experimental Pragmatics. New York: Palgrave. Kaup, Barbara, and Jana Lüdtke (). Accessing discourse referents introduced in negated phrases: Evidence for accommodation? In A. Benz and P. Kühnlein (eds.), Constraints in Discourse, –. Amsterdam: John Benjamins. Kaup, Barbara, and R. A. Zwaan (). Effects of negation and situational presence on the accessibility of text information. Journal of Experimental Psychology: Learning, Memory, and Cognition : –. Kaup, Barbara, Jana Lüdtke, and Rolf A. Zwaan (). Effects of negation, truth value, and delay on picture recognition after reading affirmative and negative sentences. In B. G. Bara, L. Barsalou, and M. Bucciarelli (eds.), Proceedings of the th Annual Conference of the Cognitive Science Society, –. Mahwah, NJ: Lawrence Erlbaum. Kaup, Barbara, Jana Lüdtke, and Rolf A. Zwaan (). Processing negated sentences with contradictory predicates: Is a door that is not open mentally closed? Journal of Pragmatics : –. Kaup, Barbara, Rolf A. Zwaan, and Jana Lüdtke (). The experiential view of language comprehension: How is negation represented? In F. Schmalhofer and C. A. Perfetti (eds.), Higher Level Language Processes in the Brain: Inference and Comprehension Processes, –. Mahwah, NJ: Erlbaum. Kaup, Barbara, Richard H. Yaxley, Carol J. Madden, R. A. Zwaan, and Jana Lüdtke (). Experiential simulations of negated text information. The Quarterly Journal of Experimental Psychology : –. Kay, Paul (). Pragmatic aspects of grammatical constructions. In Laurence R. Horn and Gregory Ward (eds.), The Handbook of Pragmatics, –. Oxford: Blackwell. Kayne, Richard S. (). French Syntax. Cambridge, MA: MIT Press. Kayne, Richard S. (). Overt vs. covert movement. Syntax : –. Kayne, Richard S. (). The Antisymmetry of Syntax. Cambridge, MA: MIT Press. Kayne, Richard S. (). Comparative syntax. Lingua : –. Keenan, Edward, and Leonard Faltz (). Boolean Semantics for Natural Language. Dordrecht: D. Reidel. Kemenade, Ans van (). Jespersen’s Cycle revisited: Formal properties of grammaticalization. In Susan Pintzuk, George Tsoulas, and Anthony Warner (eds.), Diachronic Syntax: Models and Mechanisms, –. Oxford: Oxford University Press. Kemenade, Ans van (). Secondary negation and information structure organization in the history of English. In Pierre Larrivée and Richard Ingham (eds.), The Evolution of Negation: Beyond the Jespersen Cycle, –. Berlin: Mouton de Gruyter. Kemmerer, David (). Are the motor features of verb meanings represented in the precentral motor cortices? Yes, but within the context of a flexible, multilevel architecture for conceptual knowledge. Psychonomic Bulletin & Review : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Kempson, Ruth M. (). Presupposition and the Delimitation of Semantics. Cambridge: Cambridge University Press. Kempson, Ruth M. (). More on any: Reply to Ladusaw. North-Eastern Linguistic Society : –. Kemtes, K. A., and S. Kemper (). Aging and resolution of quantifier scope effects. The Journals of Gerontology. Series B, Psychological Sciences and Social Sciences  (): –. Kendon, Adam (). Some uses of the headshake. Gesture  (): –. Kendon, Adam (). Gesture: Visible Action as Utterance. Cambridge: Cambridge University Press. Kenesei, István (). Quantifiers, negation, and focus on the left periphery in Hungarian. Lingua : –. Kennedy, Christopher (). Projecting the Adjective: The Syntax and Semantics of Gradability and Comparison. Ph.D. dissertation, University of California, Santa Cruz. Kennely, Sarah D. (). Turkish subordination: [-CP, +Tense+Case]. In A. Konrot (ed.), Modern Studies in Turkish Linguistics, –. Eskişehir: Adadolu University. Kettner, Victoria A., and Jeremy I. M. Carpendale (). Developing gestures for no and yes: Head shaking and head nodding in infancy. Gesture  (): –. Kidd, E., S. Donnelly, and M. H. Christiansen (). Individual differences in language acquisition and processing. Trends in Cognitive Sciences  (): –. Kim, A., and L. Sikos (). Conflict and surrender during sentence processing: An ERP study of syntax–semantics interaction. Brain and Language,  (–): –. Kim, Kyung J. (). Development of the concept of truth-functional negation. Developmental Psychology : –. Kim, Shin-Sook (). Intervention effects are focus effects. In Noriko Akatsuka and Susan Strauss (eds.), Japanese/Korean Linguistics, : –. Stanford, CA: CSLI. Kim, Shin-Sook, and Peter Sells (). Generalizing the immediate scope constraint on NPI licensing. In Proceedings of the Workshop on Negation and Polarity, pp. –. Kim, Young-Joo (). The acquisition of Korean. In Dan Slobin (ed.) The Crosslinguistic Study of Language Acquisition, Vol. , –. Hillsdale, NJ: Lawrence Erlbaum. Kimmelman, Vadim (). Quantification in Russian Sign Language. In Denis Paperno and Edward L. Keenan (eds.), Handbook of Quantifiers in Natural Language, Vol. , Studies in Linguistics and Philosophy , –. Berlin: Springer. King, Tracy Holloway (). Licensing negative pronominals in Georgian. In Jan Johnson, Mattew L. Juge, and Jeri L. Moxley (eds.), Proceedings of the Twenty-Second Annual Meeting of the Berkeley Linguistics Society: General Session and Parasession on the Role of Learnability in Grammatical Theory, –. Berkeley: Berkeley Linguistics Society. Kiparsky, Paul (). Linguistic universals and linguistic change. In Emmon Bach and Robert T. Harms (eds.), Universals in Linguistic Theory, –. New York: Holt, Rinehart, and Winston. Kiparsky, Paul (). ‘Elsewhere’ in phonology. In Stephen R. Anderson and Paul Kiparsky (eds.), A Festschrift for Morris Halle, –. New York: Holt, Rinehart & Winston. Kiparsky, Paul, and Cleo Condoravdi (). Tracking Jespersen’s Cycle. In Mark Janse, Brian Joseph, and Angela Ralli (eds.), Proceedings of the nd International Conference of Modern Greek Dialects and Linguistic Theory, –. Mytilene: Doukas. Kiparsky, Paul, and Carol Kiparsky (). Fact. In M. Bierwisch and K. E. Heidolph (eds.), Progress in Linguistics, –. The Hague: Mouton. Kiss, Katalin É. (). Identificational focus versus information focus. Language : –. Kiss, Katalin É. (). The evolution of functional left peripheries in Hungarian syntax. In Katalin É. Kiss (ed.), The Evolution of Functional Left Peripheries in Hungarian Syntax, –. Oxford: Oxford University Press.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Kizach, Johannes, Ken Ramshøj Christensen, and Ethan Weed (). A verbal illusion: Now in three languages. Journal of Psycholinguistic Research  (): –. Kjellmer, Göran (). Negated adjectives in modern English. Studia Neophilologica  (): –. Klamer, Marian, Ger P. Reesink, and Miriam van Staden (). East Nusantara as a linguistic area. In Pieter Muysken (ed.), From Linguistic Areas to Areal Linguistics, –. Amsterdam: Benjamins. Klima, Edward (). Negation in English. In Jerry A. Fodor and Jerrold Katz (eds.), The Structure of Language, –. Englewood Cliffs, NJ: Prentice Hall. Klima, Edward, and Ursula Bellugi (). Syntactic regularities in the speech of children. In J. Lyons and R. Wales (eds.), Psycholinguistic Papers, –. Edinburgh: Edinburgh University Press. Klinedinst, N. (). Quantified conditionals and conditional excluded middle. Journal of Semantics  (): –. Klooster, Wim (). Negative raising revisited. In J. Koster and H. van Riemsdijk (eds.), Germania et Alia: A Linguistic Festschrift for Hans den Besten. Available at . Ko, Heejeong (). Syntax of Why-in-situ: Merge into [Spec,CP] in the overt syntax. Natural Language and Linguistic Theory : –. Koeneman, Olaf, and Hedde Zeijlstra (). Introducing Syntax. Cambridge: Cambridge University Press. König, Ekkehard (). Temporal and non-temporal uses of schon and noch in German. Linguistics and Philosophy : –. König, Ekkehard (). The Meaning of Focus Particles: A Comparative Perspective. London: Routledge. König, Ekkehard (). Syntax and semantics of additive focus markers from a cross-linguistic perspective: A tentative assessment of the state of the art. In A. de Cesare and C. Adorno (eds.), Focus on Additivity: Adverbial Modifiers in Romance, Germanic and Slavic Languages, –. Amsterdam: John Benjamins. König, Ekkehard, and Elizabeth Closs Traugott (). Divergence and apparent convergence in the development of yet and still. In Berkeley Linguistic Society, –. Berkeley. Kornfilt, Jaklin (). Turkish. London: Routledge. Kortmann, Bernd, and Kerstin Lunkenheimer (eds.) (). Electronic Atlas of World Englishes (Feature ). URL: . Koster, E. H. W., E. Rassin, G. Crombez, and G. W. B. Näring (). The paradoxical effects of suppressing anxious thoughts during imminent threat. Behaviour Research and Therapy : –. Kotek, Hadas (). Composing Questions. Ph.D. dissertation, Massachusetts Institute of Technology, Cambridge, MA. Kotilainen, Lari (). Konstruktioiden dynamiikkaa. Ph.D. dissertation, University of Helsinki. Kotz, Sonja A. (). A critical review of ERP and fMRI evidence on L syntactic processing. Brain and Language : –. Kounios, J., and P. J. Holcomb (). Structure and process in semantic memory: Evidence from event-related brain potentials and reaction times. Journal of Experimental Psychology: General : –. Kovács, Ágnes M. (). Extracting regularities from noise: Do infants encode patterns based on same and different relations? Language Learning  (s): –. Kraak, Albert (). Negatieve zinnen. Hilversum: W. de Haan. Kramer, Raija (ed.) (forthcoming). The Expression of Phasal Polarity in African Languages. Berlin: DeGruyter. Kramer, Ruth, and Kyle Rawlins (). Polarity particles: An ellipsis account. In Suzi Lima, Kevin Mullin, and Brian Smith (eds.), Proceedings of NELS : –. Amherst, MA: GLSA.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Krasnoukhova, Olga, and Johan van der Auwera (under review). Negative ma in South America with a focus on its standard negative postverbal ma in Panoan, Takanan and Tukanoan. In Mily Crevels, Hein van der Voort, and Matti Miestamo (eds.), Negation in Amazonian Languages (Studies in Diversity Linguistics series). Berlin: Language Science Press. Kratzer, Angelika (). The notional category of modality. In Hans J. Eikemeyer and Hannes Rieser (eds.), Words, Worlds, and Contexts, –. Berlin: de Gruyter. Kratzer, Angelika (). An investigation of the lumps of thought. Linguistics and Philosophy  (): –. Kratzer, Angelika () Modality. In Dieter Wunderlich and Arnim von Stechow (eds.), Semantics: An International Handbook of Contemporary Research, –. Berlin: de Gruyter. Kratzer, Angelika (). Severing the external argument from its verb. In Johan Rooryck and Laurie Zaring (eds.), Phrase Structure and the Lexicon. Dordrecht: Kluwer, –. Kratzer, Angelika, and Junko Shimoyama (). Indeterminate pronouns: The view from Japanese. In Y. Otsu (ed.), Proceedings of the Third Tokyo Conference on Psycholinguistics, –. Tokyo: Hituzi Syobo. Krifka, Manfred (). Some remarks on polarity items. In D. Zaefferer (ed.), Semantic Universals and Universal Semantics, GRASS, –. Berlin: Foris. Krifka, Manfred (). The semantics and pragmatics of weak and strong polarity items in assertions. In Proceedings of Semantics and Linguistic Theory IV, –. Krifka, Manfred (). The semantics and pragmatics of polarity items. Linguistic Analysis : –. Krifka, Manfred (). Pragmatic strengthening in plural predications and donkey sentences. In Teresa Galloway and Justin Spence (eds.), Semantics and Linguistic Theory (SALT) , –. Ithaca, NY: CLC Publications. Krifka, Manfred (). Scope inversion under rise–fall contour in German. Linguistic Inquiry  (): –. Krifka, Manfred (). Association with focus phrases. In V. Molnar and S. Winkler (eds.), The Architecture of Focus, –. Berlin: Mouton de Gruyter. Krifka, Manfred (a). Basic notions of information structure. In C. Féry, G. Fanselow, and Manfred Krifka (eds.), Interdisciplinary Studies on Information Structure : –. Krifka, Manfred (b). Negated antonyms: Creating and filling the gap. In U. Sauerland and P. Stateva (eds.), Presupposition and Implicature in Compositional Semantics, –. Houndmills: Palgrave Macmillan. Krifka, Manfred (). How to interpret “expletive” negation under bevor in German. In T. Hanneforth and G. Fanselow (eds.), Language and Logos: Studies in Theoretical and Computational Linguistics. Akademie Verlag (Studia Grammatica ). Krifka, Manfred (). Bias in commitment space semantics: Declarative questions, negated questions, and question tags. In Sarah D’Antonio, Mary Moroney, and Carol Rose Little (eds.), Proceedings of Semantics and Linguistic Theory , –. Washington, DC: LSA Open Journal Systems. Krifka, Manfred (). Negative polarity questions as denegations of assertions. In Chungmin Lee, Ferenc Kiefer, and Manfred Krifka (eds.), Contrastiveness in Information Structure, Alternatives and Scalar Implicatures, –. Berlin: Springer. (Circulated as ms. since .) Križ, Manuel (). Homogeneity, Non-Maximality, and all. Journal of Semantics : –. Kroch, Anthony (). Reflexes of grammar in patterns of language change. Language Variation and Change : –. Kroch, Anthony (). Morphosyntactic variation. In K. Beals et al. (eds.), Proceedings of the th Annual Meeting of the Chicago Linguistics Society (Parasession on Variation and Linguistic Theory.), Vol , –. Kroch, Anthony (). Amount quantification, referentiality, and long wh-movement. In University of Pennsylvania Working Papers in Linguistics : –. Philadelphia: University of Pennsylvania.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Kroch, Anthony, and Ann Taylor (). The Penn-Helsinki Parsed Corpus of Middle English Prose (nd edn.). Philadelphia: Department of Linguistics, University of Pennsylvania. Kroch, Anthony, Beatrice Santorini, and Lauren Delfs (eds.) (). Penn-Helsinki Parsed Corpus of Early Modern English. Philadelphia: Department of Linguistics, University of Pennsylvania. Kroeger, Paul (). External negation in Malay/Indonesian. Language  (): –. Kroll, J. F., and A. M. B. DeGroot (). Handbook of Bilingualism: Psycholinguistic Approaches. Oxford: Oxford University Press. Kroskrity, Paul V. (). Negation and subordination in Arizona Tewa: Discourse pragmatics influencing syntax. International Journal of American Linguistics : –. Krupenye, Christopher, Fumihiro Kano, Satoshi Hirata, Josep Call, and Michael Tomasello (). Great apes anticipate that other individuals will act according to false beliefs. Science : –. Kügler, Frank (). Do we know the answer?—Variation in yes-no question intonation. In Susann Fischer, Ruben van de Vijver, and Ralf Vogel (eds.), Experimental Studies in Linguistics : –. Potsdam: Universitätsverlag Potsdam. Kulakova, E., and M. S. Nieuwland (). Understanding counterfactuality: A review of experimental evidence for the dual meaning of counterfactuals. Language and Linguistics Compass : –. Kumar, Rajesh (). Negation and Licensing of Negative Polarity Items in Hindi Syntax. London: Routledge. Kuno, Masakazu (). Negation, focus, and negative concord in Japanese. Toronto Working Papers in Linguistics : –. Kuno, Susumu, and Ken-ichi Takami (). Remarks on negative islands. Linguistic Inquiry : –. Kuperberg, G. R. (). Neural mechanisms of language comprehension: Challenges to syntax. Brain Research : –. Kuperberg, G. R., D. A. Kreher, T. Sitnikova, D. N. Caplan, and P. J. Holcomb (). The role of animacy and thematic relationships in processing active English sentences: Evidence from event-related potentials. Brain and Language  (): –. Kurcz, Ida (). The influence of grammatical habits on sentence recognition. In G. B. Flores D’Arcais and W. J. M. Levelt (eds.), Advances in Psycholinguistics, –. Amsterdam: North Holland. Kuroda, S.-Y. (). The categorical and the thetic judgment (evidence from Japanese). Foundations of Language : –. Kurtzman, H. S., and M. C. MacDonald (). Resolution of quantifier scope ambiguities. Cognition  (): –. Kush, D., J. Lidz, and C. Phillips (). Relation-sensitive retrieval: Evidence from bound variable pronouns. Journal of Memory and Language : –. Kutas, Marta (). One lesson learned: Frame language processing—literal and figurative—as a human brain function. Metaphor and Symbol : –. Kutas, Marta, and Kara D. Federmeier (). Thirty years and counting: Finding meaning in the N component of the event-related brain potential (ERP). Annual Review of Psychology : –. Kutas, Marta, and Steven A. Hillyard (). Reading senseless sentences: Brain potentials reflect semantic incongruity. Science  (): –. Kutas, Marta, and Steven A. Hillyard (). Event-related brain potentials to grammatical errors and semantic anomalies. Mem. Cogn. : –. Labelle, Marie, and M.Teresa Espinal (). Diachronic changes in negative expressions: The case of French. Lingua : –. Labov, William (). Contraction, deletion, and the inherent variability of the English copula. Language : –. Labov, William (a). Language of the Inner City. Philadelphia: University of Pennsylvania Press. Labov, William (b). Negative attraction and negative concord in English grammar. Language : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Labov, William, Paul Cohen, Clarence Robins, and John Lewis (). A Study of the Nonstandard English of Negro and Puerto Rican Speakers in New York City, Final Report. Cooperative Research Project No. , United States Office of Education. LaBrum, Rebecca W. (). Conditions on Double Negation in the History of English with Comparison to Similar Developments in German. Ph.D. thesis, Stanford University, Stanford, CA. Ladd, D. Robert (). The Structure of Intonational Meaning. Bloomington: Indiana University Press. Ladd, D. Robert (). A first look at the semantics and pragmatics of negative questions and tag questions. In Roberta A. Hendrick, Carrie S. Masek, and Mary Frances Miller (eds.), Papers from the th Regional Meeting of the Chicago Linguistic Society, –. Chicago: CLS. Ladusaw, William A. (). Polarity sensitivity as inherent scope relations. Ph.D. dissertation University of Texas. Ladusaw, William A. (a). On the notion of affective in the analysis of negative polarity items. Journal of Linguistics : –. Ladusaw, William A. (b). Polarity Sensitivity as Inherent Scope Relations. New York: Garland. Ladusaw, William A. (). Expressing negation. In Chris Barker and David Dowty (eds.), Semantics and Linguistic Theory (SALT) : –. Ithaca, NY: Cornell University. Ladusaw, William A. (). Negation, indefinites, and the Jespersen cycle. In Joshua S. Guenter, Barbara A. Kaiser, and Cheryl C. Zoll (eds.), Proceedings of the th Annual Meeting of the Berkeley Linguistics Society, –. Berkeley: Berkeley Linguistics Society. Ladusaw, William A. (). Negation and polarity items. In Shalom Lappin (ed.), The Handbook of Contemporary Semantic Theory, –. Oxford: Blackwell. Ladusaw, William A. (). Framing the issues: The biasing effect of polarity items in questions. Talk at First International Workshop on Current Research in the Semantics-Pragmatics Interface. LSA summer institute. Lafkioui, Mena B. (). Negation, grammaticalization and language change in North Africa: The case of the negator NEG __ *bu. In Giorgio Francesco Arcodia, Frederica Da Milano, Gabriele Iannàccaro, and Paolo Zublena (eds.), Tilelli. Scritti in onore di Vermondo Brugnatelli, –. Cesena/Rome: Caissa Italia S.c.a.r.l. Lafkioui, Mena B., and Vermondo Brugnatelli (forthcoming). Negation in Berber: Variation, evolution and typology. Linguistics. Lahiri, Utpal (). Focus and negative polarity in Hindi. Natural Language Semantics  (): –. Lahiri, Utpal (). The semantics and pragmatics of some scalar expressions in Spanish. Anuario del Seminario de Filología Vasca Julio de Urquijo : –. Laka, Itziar (). Negation in Syntax: On the Nature of Functional Categories and Projections. Ph.D. dissertation, MIT. (published  Garland). Laka, Itziar (). Sentence negation in Basque. ASJU: Anuario del Seminario de Filologia Vasca “Julio de Urquijo” .: –. Laka, Itziar (). On the Syntax of Negation. Outstanding Dissertations in Linguistics Series. New York and London: Garland Publishing Co. Lakoff, George (a). A note on vagueness and ambiguity. Linguistic Inquiry  (): –. Lakoff, George (b). Pronominalization, negation, and the analysis of adverbs. In R. Jacobs and P. Rosenbaum (eds.), Readings in English Transformational Grammar, –. Waltham: Ginn. Lakoff, Robin (a). A syntactic argument for negative transportation. CLS : –. Lakoff, Robin (b). Some reasons why there can’t be any some–any rule. Language : –. Lakoff, Robin (). If ’s, and’s and but’s about conjunction. In Charles J. Fillmore and D. Terrence Langendoen (eds.), Studies in Linguistic Semantics, –. New York: Irvington. Lakshmanan, Usha, and Simona Maciukaite (). The syntax and pragmatics of negation in child SLA. Proceedings of the Pragmatics and Language Learning Conference : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Lambon-Ralph, Matthew A. (). Neurocognitive insights on conceptual knowledge and its breakdown. Philosophical Transactions of the Royal Society of London B: Biological Sciences : . Lambrecht, Knud (). Information Structure and Sentence Form: Topic, Focus, and the Mental Representation of Discourse Referents. Cambridge: Cambridge University Press. Landin, David J. (). An outline of the syntactic structure of Karitiâna sentences. In Robert A. Dooley (ed.), Estudos sobre linguas tupi do Brasil, –. Brasilia: Summer Institute of Linguistics. Langacker, Ronald W. (). Foundations of Cognitive Grammar. Stanford, CA: Stanford University Press. Langbauer, W. R., K. B. Payne, R. A. Charif, L. Rapaport, and F. Osborn (). African elephants respond to distant playbacks of low-frequency conspecific calls. Journal of Experimental Biology  (): –. Langendoen, D. T., and H. Savin (). The projection problem for presuppositions. In C. Fillmore and D. T. Langendoen (eds.), Studies in Linguistic Semantics. New York: Holt, Rinehart, and Winston. Lardiere, Donna (). Dissociating syntax from morphology in a divergent L end-state grammar. Second Language Research  (): –. Larrivée, Pierre (). The pragmatic motifs of the Jespersen Cycle: Default, activation and the history of negation in French. Lingua  (): –. Larrivée, Pierre (a). La définition de la négation métalinguistique. In Jacqueline Bacha, Ammar Azouzi, and Khaled Salddem (eds.), La négation en discours, –. Sousse: Université de Sousse. Larrivée, Pierre (b). The role of pragmatics for grammatical change: The case of French preverbal non. Journal of Pragmatics  (): –. Larrivée, Pierre (a). The syntax of pragmatics: The case of presuppositional Negatives. Syntaxe et Sémantique : –. Larrivée, Pierre (b). L’unidirectionalité irréversible du changement linguistique comme conséquence de l’acquisition? Le cas d’expressions négatives déliquescentes en français ancien. In Franck Neveu, Peter Blumenthal, Linda Hriba, Annette Gerstenberg, Judith Meinschaefer, and Sophie Prévost (eds.), Congrès Mondial de Linguistique Française , –. SHS Web of Conferences. Larrivée, Pierre (a). The markedness of double negation. In P. Larrivée and C. Lee (eds.), Negation and Polarity: Experimental Perspectives, Language, Cognition, and Mind , –. Springer. Larrivée, Pierre (b). The pragmatics of marked configurations: Negative doubling in French. Journal of Pragmatics : –. Larrivée, Pierre (). Metalinguistic negation from an informational perspective. Glossa: A Journal of General Linguistics  (): . –. Larrivée, Pierre (). Historical pragmatics, explicit activation and wh in situ in French. In Ingo Feldhausen, Martin Elsig, Imme Kuchenbrandt, and Mareike Neuhaus (eds.), Romance Languages and Linguistic Theory : Selected Papers from ‘Going Romance’ , –, Frankfurt, Amsterdam: Benjamins. Larrivée, Pierre, and Richard Ingham (eds.) (). The Evolution of Negation: Beyond the Jespersen Cycle. Berlin: Mouton de Gruyter. Lasersohn, Peter (). Context dependence, disagreement, and predicates of personal taste. Linguistics and Philosophy : –. Lasnik, Howard (). Analyses of Negation in English. Ph.D. Dissertation, MIT. Laughren, Mary, Rob Pensalfini, and Tom Mylne (). Accounting for verb-initial order in an Australian language. In Andrew Carnie, Heidi Harley, and Sheila Ann Dooley (eds.), Verb First: On the Syntax of Verb-Initial Languages, –. Amsterdam: Benjamins. Lazard, Gilbert, and Louise Peltzer (). La négation en tahitien. In Hovdhaugen and Mosel (eds.), Negation in Oceanic Languages, –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Leavens, David A., Autumn B. Hostetter, Michael J. Wesley, and William D. Hopkins (). Tactical use of unimodal and bimodal communication by chimpanzees, Pan troglodytes. Animal Behaviour : –. Lechner, Winfried (). An interpretive effect of head movement. In Mara Frascarelli (ed.), Phases of Interpretation, –. Berlin: Mouton de Gruyter. Lee, Chungmin (). Types of NPIs and nonveridicality in Korean and other languages. In G. Storto (ed.), Syntax at Sunset  (UCLA Working Papers in Linguistics ), –. Los Angeles: University of California. Lee, Chungmin (). How metalinguistic negation differs from descriptive negation: ERP evidence. Proceedings of EAPCogSCI , –. AISC. Lee, Chungmin (). Metalinguistically negated versus descriptively negated adverbials: ERP and other evidence. In Pierre Larrivé and Chungmin Lee (eds.), Negation and Polarity: Experimental Perspectives, –. Berlin: Springer. Lee, Ernest W. (). Bipartite negatives in Chamic. Mon-Khmer Studies : –. Lee, Jung-Hyuck (). Korean EVEN: -to versus -(i)lato. Ms. University of Chicago. Lee, Jung-Hyuck (). Korean EVEN: -to versus -(i)lato and the nonveridicality hypothesis. Harvard Working Papers on Japanese and Korean Linguistics. Lee, Sung-Bum, and Seung Jin Hong (). An experimental study of neg-raising inferences in Korean. In P. Larrivée and C. Lee (eds.), Negation and Polarity: Experimental Perspectives, –. Berlin: Springer. Lee, Sunyoung () Interpreting Ambiguity in First and Second Language Processing: Universal Quantifiers and Negation. Unpublished Ph.D. dissertation, University of Hawai’i, Manoa, HI, USA. Lee, Young-Suk, and Laurence R. Horn (). Any as indefinite + even. Ms., Yale University. Leech, Geoffrey, Stig Johansson, Roger Garside, and Knut Hofland (–). The LOB Corpus, POStagged version. University of Oslo. Lehmann, Winfred P. (). Proto-Indo-European Syntax. Austin, TX: University of Texas Press. Lehrer, Adrienne J. (). Markedness and antonymy. Journal of Linguistics  (): –. Lehrer, Adrienne J. (). Paradigmatic relations of exclusion and opposition I: Gradable antonymy and complementarity. In D. A. Cruse, F. Hundsnurscher, M. Job, and P.-R. Lutzeier (eds.), Lexicology: An International Handbook on the Nature and Structure of Words and Vocabularies, Vol. , –. Berlin: de Gruyter. Leivada, Evelina, Elena Papadopoulou, Maria Kambanaros, and Kleanthes Grohmann (). The influence of bilectalism and nonstandardization on the perception of native grammatical variants. Frontiers in Psychology . Article . Lerner, Jean-Yves, and Wolfgang Sternefeld (). Zum Skopus der Negation im komplexen Satzen des Deutschen. Zeitschrift für Sprachwissenschaft : –. Leslau, Wolf (). Amharic textbook. Wiesbaden: Otto Harassowitz. Levine, Robert (). The modal need VP gap (non)anomaly. In E. Csipak, R. Eckardt, M. Liu, and M. Sailer (eds.), Beyond Any and Ever: New Explorations in Negative Polarity Sensitivity, –. Berlin: Mouton de Gruyter. Levine, W. H., and J. A. Hagaman (). Negated concepts interfere with anaphor resolution. Intercultural Pragmatics : –. Levinson, Stephen C. (). Presumptive Meanings: The Theory of Generalized Conversational Implicature. Cambridge, MA: MIT Press. Levinson, Stephen C., and E. Annamalai (). Why presuppositions aren’t conventional. In R. N. Srivastava (ed.), Language and Text: Studies in Honour of Ashok R. Kelkar, –. Delhi: Kalinga. Levy, R., K. Bicknell, T. Slattery, and K. Rayner (). Eye movement evidence that readers maintain and act on uncertainty about past linguistic input. Proceedings of the National Academy of Sciences  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Lewis, David (). Adverbs of quantification. In Edward L. Keenan (ed.), Formal Semantics of Natural Language, –. Cambridge: Cambridge University Press. Lewis, David (). Scorekeeping in a language game. In R. Bäuerle, U. Egli, and A. von Stechow (eds.), Semantics from Different Points of View, –. Berlin: Springer. Reprinted in Paul H. Portner and Barbara H. Partee () (eds.), Formal Semantics: The Essential Readings, –. Oxford: Blackwell. Lewis, R. L., and S. Vasishth (). An activation-based model of sentence processing as skilled memory retrieval. Cognitive Science  (): –. Li, Fei-Fei, Joan Borràs-Comes, and M.Teresa Espinal (). Single negation interpretations in sentences with multiple negative expressions in Mandarin Chinese: An experimental investigation. Lingua –: –. Li, Fei-Fei, Santiago González-Fuente, Pilar Prieto, and M.Teresa Espinal (). Is Mandarin Chinese a truth-based language? Rejecting responses to negative assertions and questions. Frontiers in Psychology : . Li, Wei (). Dimensions of bilingualism. In Wei Li (ed.), The Bilingualism Reader, –. Hove: Psychology Press. Lidz, J., and J. Musolino (). Children’s command of quantification. Cognition : –. Liebal, Katja, Josep Call, Michael Tomasello, and Simone Pika (). To move or not to move: How apes adjust to the attentional state of others. Interaction Studies : –. Lightfoot, David W. (). How to Set Parameters: Arguments from Language Change. Cambridge, MA: MIT Press. Lightfoot, David W. (). The Development of Language: Acquisition, Change and Evolution. Oxford: Blackwell. Lillo-Martin, Diane, and Ronice Müller de Quadros (). Focus constructions in American Sign Language and Brazilian Sign Language. In Josep Quer (ed.), Signs of the Time. Selected Papers from TISLR , –. Hamburg: Signum Press. Lin, Jing (). Acquiring Negative Polarity Items. Ph.D. dissertation University of Amsterdam. Lin, Jing, Fred Weerman, and Hedde Zeijlstra (). Emerging NPIs: The acquisition of Dutch hoeven (‘need’). The Linguistic Review : –. Lindholm, James (). Negative raising and sentence pronominalization. CLS : –. Lindholm, Kathryn, and Amado Padilla (). Language mixing in bilingual children. Journal of Child Language : –. Linebarger, Marcia Christine (). The Grammar of Negative Polarity. Ph.D. Dissertation, MIT. Linebarger, Marcia Christine (). Negative polarity and grammatical representation. Linguistics and Philosophy : –. Lippi-Green, Rosina (). English with an Accent: Language, Ideology, and Discrimination in the United States. London and New York: Routledge. Lipski, John M. (a). Language switching constraints: more than syntax? Data from Media Lengua. Bilingualism, Language and Cognition : –. Lipski, John M. (b). Palenquero vs. Spanish negation: Separate but equal? Lingua : –. Lipski, John M. (). Can NEG placement have negative consequences (for efficient processing)? A bilingual test case. Applied Psycholinguistics : –. Liuzza, Marco Tullio, Matteo Candidi, and Salvatore Maria Aglioti (). Do not resonate with actions: Sentence polarity modulates corticospinal excitability during action-related sentence reading. PloS One : e. Ljung, Magnus (). Swearing: A Cross-Cultural Linguistic Study. Houndmills: Palgrave Macmillan. Löbner, Sebastian (). Definites. Jourmal of Semantics : –. Löbner, Sebastian (). Quantification as a major module of natural language semantics. In Jeroen Groenendijk, Dick de Jongh, and Martin Stokhof (eds.), Studies in Discourse Representation Theory and the Theory of Generalized Quantifiers, –. Dordrecht: Foris.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Löbner, Sebastian (). German schon—erst—noch: An integrated analysis. Linguistics and Philosophy : –. Löbner, Sebastian (). Wahr neben Falsch: Duale Operatoren als die Quantoren naturlicher Sprache. Tübingen: Max Niemeyer. Löbner, Sebastian (). Why German schon and noch are still duals: A reply to van der Auwera. Linguistics and Philosophy : –. Löbner, Sebastian (). Polarity in natural language: Predication, quantification and negation in particular and characterizing sentences. Linguistics & Philosophy : –. Lohndal, Terje (). Phrase Structure and Argument Structure: A Case Study of the Syntax–Semantics Interface. Oxford Studies in Theoretical Linguistics. Oxford: Oxford University Press. Longobardi, Giuseppe (). Island effects and parasitic constructions. In Jim Huang and Robert May (eds.), Logical Structure & Linguistic Structure, –. Dordrecht: Kluwer. Longobardi, Giuseppe (). Theory and experiment in parametric minimalism: The case of Romance negation. In Rob Pensalfini, Myfany Turpin, and Diana Guillemin (eds.), Language Description Informed by Theory, –. Amsterdam: Benjamins. Lønning, Jan Tore (). Mass terms and quantification. Linguistics & Philosophy  (): –. Lonzi, L., and C. Luzzatti (). Relevance of adverb distribution for the analysis of sentence representation in agrammatic patients. Brain and Language  (): –. López, Luis (). Bilingual grammar: Toward an integrated model. Ms., UIC. Lucas, Christopher (). Jespersen’s Cycle in Arabic and Berber. Transactions of the Philological Society  (): –. Lucas, Christopher (). The development of negation in Arabic and Afro-Asiatic. Ph.D. dissertation, University of Cambridge. Lucas, Christopher (). Negation in the history of Arabic and Afro-Asiatic. In David Willis, Christopher Lucas, and Anne Breitbarth (eds.), The History of Negation in the Languages of Europe and the Mediterranean, Vol. : Case Studies, –. Oxford: Oxford University Press. Lucas, Christopher, and David Willis (). Never again: The multiple grammaticalization of never as a marker of negation in English. English Language and Linguistics  (): –. Lüdtke, J., and B. Kaup (). Context effects when reading negative and affirmative sentences. In R. Sun (ed.), Proceedings of the th Annual Conference of the Cognitive Science Society, –. Mahwah, NJ: Lawrence Erlbaum. Lüdtke, J., C. K. Friedrich, M. De Filippis, and B. Kaup. (). Event-related potential correlates of negation in a sentence–picture verification paradigm. Journal of Cognitive Neuroscience : –. Lukasiewicz, Jan (). Philosophical remarks on many-valued systems of propositional logics. In S. McCall (ed.), Polish Logic –, –. Oxford: Clarendon. Luo, Yuyan, and Renée Baillargeon (). When the ordinary seems unexpected: Evidence for incremental physical knowledge in young infants. Cognition : –. Lycan, William G. (). The Myth of the “Projection Problem for Presupposition.” Philosophical Topics  (): –. Lyons, John (). Logic, Vol. . Cambridge: Cambridge University Press. McCawley, James D. (). Contrastive negation and metalinguistic negation. Proceedings of CLS . Part II: The Parasession on Negation, –. Chicago: Chicago Linguistic Society. McCawley, James D. (a). The focus and scope of only. In B. Partee and P. Sgall (eds.), Discourse and Meaning: Papers in Honor of Eva Hajicová, –. Amsterdam: John Benjamins. McCawley, James D. (b). The Syntactic Phenomena of English. Chicago: University of Chicago Press. McCloskey, James (). Transformational Syntax and Model Theoretic Semantics: A Case-Study in Modern Irish. Dordrecht: Reidel.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





McCloskey, James (). Constraints on syntactic processes. In Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld, and Theo Vennemann (eds.), Syntax: An International Handbook of Contemporary Research, Vol. , –. Berlin: Mouton de Gruyter. McCloskey, James (). Polarity and case licensing: The cartography of the inflectional layer in Irish. Talk presented at GIST : Cartographic Structures and Beyond. Ghent University. McCloskey, James (). Ellipsis, polarity, and the cartography of verb-initial orders in Irish. Ms. UCSC. MacDonald, Maryellen C., and Marcel A. Just. (). Changes in activation levels with negation. Journal of Experimental Psychology: Learning, Memory, and Cognition  (): . Machery, Edouard (). The amodal brain and the offloading hypothesis. Psychonomic Bulletin & Review : –. McKee, Rachel (). Aspects of interrogatives and negation in New Zealand Sign Language. In Ulrike Zeshan (ed.), Interrogative and Negative Constructions in Sign Languages, –. Nijmegen: Ishara Press. Magri, Giorgio (). An account for the homogeneity effects triggered by plural definites and conjunction based on double strengthening. In Salvatore Pistoia Reda (ed.), Semantics, Pragmatics and the Case of Scalar Implicatures, –. Basingstoke: Palgrave Macmillan. Maguire, M. J., M. R. Brier, P. S. Moore, T. C. Ferree, D. Ray, S. Mostofsky, J. Hart Jr, and M. A. Kraut (). The influence of perceptual and semantic categorization on inhibitory processing as measured by the N–P response. Brain Cogn : –. Mahajan, Anoop (). Agreement and agreement phrases. MIT Working Papers in Linguistics : –. Mahon, Bradford Z., and Alfonso Caramazza (). A critical look at the embodied cognition hypothesis and a new proposal for grounding conceptual content. Journal of Physiology-Paris  (–): –. Makharoblidze, Tamar, and Roland Pfau (). A negation-tense interaction in Georgian Sign Language. Sign Language & Linguistics  (): –. Makuuchi, M., Y. Grodzinsky, K. Amunts, A. Santi, and A. D. Friederici (). Processing noncanonical sentences in Broca’s region: Reflections of movement distance and type. Cerebral Cortex  (): –. Malamud, Sophia A. (). The meaning of plural definites: A decision-theoretic approach. Semantics & Pragmatics : –. Malamud, Sophia A., and Tamina Stephenson (). Three ways to avoid commitments: Declarative force modifiers in the conversational scoreboard. Journal of Semantics : –. Manzini, Maria Rita, and Leonardo M. Savoia (). The syntax of middle-reflexive and object clitics: A case of parametrization in arbëresh dialects. In Matteo Mandalà (ed.), Studi in onore di Luigi Marlekaj, –. Bari: Adriatica. Manzini, Maria Rita, and Leonardo M. Savoia (). Grammatical Categories. Cambridge: Cambridge University Press. Manzotti, E., and A. Rigamonti (). La negazione. In D. Renzi et al. (eds.), Grande Grammatica Italiana di Consultazione, Vols. –. Bologna: Il Mulino. Marques, Teresa (). Truth and the ambiguity of negation. In Erich Rast and Luiz Carlos Baptista (eds.), Meaning and Context, –. Bern: Peter Lang. Marsh, Heidi L., and Suzanne E. MacDonald (). Information seeking by orangutans: A generalized search strategy? Animal Cognition : –. Martineau, France, and Viviane Déprez (). Pas rien/Pas aucun en français classique: Variation dialectale et historique. Langue francaise : –. Martineau, France, and Raymond Mougeon (). A sociolinguistic study of the origins of nedeletion in European and Quebec French. Language : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Martins, Ana Maria (). Polarity Items in Romance: Underspecification and lexical change. In Susan Pintzuk, George Tsoulas, and Anthony Warner (eds.), Diachronic Syntax: Models and Mechanisms, –. Oxford: Oxford University Press. Martins, Ana Maria (). Emphatic affirmatio and polarity: Contrasting European Portuguese with Brazilian Portuguese, Spanish, Catalan, Galician. In Jenny Doetjes and Paz Gonzalez (eds.), –. Romance Languages and Linguistic Theory. Dordrecht: John Benjamins. Martins, Ana Maria (). Negação metalinguística (lá, cá e agora). In Ana Maria Brito, Fátima Silva, João Veloso, and Alexandra Fiéis (eds.), Actas do XXV Encontro Nacional da Associação Portuguesa de Linguística, –. Porto: Associação Portuguesa de Linguística. Martins, Ana Maria (). Deictic locatives, emphasis and metalinguistic negation. In Charlotte Galves, Sonia Cyrino, Ruth Lopes, Filomena Sandalo, and Juanito Avelar (eds.), Parameter Theory and Linguistic Change, –. Oxford/New York: Oxford University Press. Martins, Ana Maria (). Emphatic polarity in European Portuguese and beyond. Lingua : –. Martins, Ana Maria (). How much syntax is there in metalinguistic negation? Natural Language and Linguistic Theory  (): –. Martins, Ana Maria (). Negation and NPI composition inside DP. In Theresa Biberauer and George Walkden (eds.), Syntax over Time, –. Oxford: Oxford University Press. Mason, Robert A., and Marcel Adam Just (). Neuroimaging contributions to the understanding of discourse processes. In Handbook of Psycholinguistics (nd edn.), –. Amsterdam: Elsevier. Massam, Diane (). VSO and VOS: Aspects of Niuean word order. In Andrew Carnie and Eithne Guilfoyle (eds.), The Syntax of Verb Initial Languages, –. Oxford: Oxford University Press. Master, Alfred (). The zero negative in Dravidian. Transactions of the Philological Society : –. Mateo, Jill M. (). The development of alarm-call response behaviour in free-living juvenile Belding’s ground squirrels. Animal Behaviour  (): –. Mates, Benson (). Stoic Logic. Berkeley: University of California Press. Matthewson, Lisa (). Presuppositions and cross-linguistic variation. In Proceedings of NELS . Amherst, MA: University of Massachusetts. Matushansky, Ora (). Head movement in linguistic theory. Linguistic Inquiry  (): –. May, Robert (). Logical Form: Its Structure and Derivation. Cambridge, MA: MIT Press. Mayo, R., Y. Schul, Y., and E. Burnstein (). “I am not guilty” vs “I am innocent”: Successful negation may depend on the schema used for its encoding. Journal of Experimental Social Psychology : –. Mayo, R., Y. Schul, and M. Rosenthal (). If you negate, you may forget: Negated repetitions impair memory compared with affirmative repetitions. Journal of Experimental Psychology: General : –. Mayr, Clemens (). Implicatures of modified numerals. In Ivano Caponigro and Carlo Cecchetto (eds.), From Grammar to Meaning: The Spontaneous Logicality of Language, –. Cambridge: Cambridge University Press. Mayr, Clemens (). Intervention effects and addivity. Journal of Semantics : –. Mayr, Clemens, and Benjamin Spector (). Not too strong! Generalizing the Scope Economy Condition. In Martin Prinzhorn, Viola Schmitt, and Sarah Zöbel (eds.), Proceedings of Sinn und Bedeutung : –. Mayr, Clemens, and Benjamin Spector (). Generalized Scope Economy—not too strong! Ms., Zentrum für Allgemeine Sprachwissenschaft and Institut Jean Nicod. Mazzon, Gabriella (). A History of English Negation. Harlow: Pearson Longman. Meibauer, Jörg (). Sentence mood, lexical category filling, and non-propositional nicht in German. Linguistische Berichte : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Meillet, Antoine (). L’évolution des formes grammaticales. Scientia  (): –. (Reprinted in Antoine Meillet (). Linguistique historique et linguistique générale, –. Paris: Champion.) Meir, Irit (). Question and negation in Israeli Sign Language. Sign Language & Linguistics  (): –. Meisel, Jürgen (). Early differentiation of languages in bilingual children. In Kenneth Hyltenstam and Loraine Obler (eds.). Bilingualism Across the Lifespan. Cambridge: Cambridge University Press. Meisel, Jürgen (). The acquisition of the syntax of negation in French and German: Contrasting first and second language development. Second Language Research  (): –. Melis, Alicia P., Josep Call, and Michael Tomasello (). Chimpanzees (Pan troglodytes) conceal visual and auditory information from others. Journal of Comparative Psychology : –. Mellet, Sylvie (). L’alternance ne/non en latin classique. L’information grammaticale : –. Mendia, J. A., E. Poole, and E. Dillon (). Spurious NPI licensing and exhaustification. Proceedings of SALT : –. Merchant, Jason (). The Syntax of Silence. Oxford: Oxford University Press. Merchant, Jason (). Remarks on stripping. Ms., University of Chicago. . Last accessed January , . Merchant, Jason (). Fragments and ellipsis. Linguistics and Philosophy  (): –. Merchant, Jason (). Why no(t)? Style  (–): –. Merchant, Jason (). Three kinds of ellipsis: Syntactic, semantic, pragmatic? In François Recanati, Isidora Stojanovic, and Neftail Villanueva (eds.), Context-Dependence, Perspective, and Relativity, –. Berlin: Mouton de Gruyter. Merchant, Jason (). Polarity items under ellipsis. In Lisa Lai-Shen Cheng and Norbert Corver (eds.), Diagnosing Syntax, –. Oxford: Oxford University Press. Merchant, Jason (). The Afrikaans final negative particle as a negative isotopic, VP-level clitic. Ms. University of Chicago. Plenary paper. st Comparative Germanic Syntax Workshop. University of Stellenbosch. Merchant, Jason (). Ellipsis: A survey of analytical approaches. In Jeroen van Craenenbroeck and Tanja Temmerman (eds.), The Oxford Handbook of Ellipsis. Oxford: Oxford University Press. Merin, Arthur (). Algebra of elementary social acts. In S. L. Tsohatzidis (ed.), Foundations of Speech Act Theory: Philosophical and Linguistic Perspectives, –. New York: Routledge. Mettouchi, Amina (). The system of negation in Berber. In Cyffer et al. (eds.), Negation Patterns in West African Languages and Beyond, –. Michael, Lev (). Negation in Nanti. In Lev Michael and Tania Granadillo (eds.), Negation in Arawak Languages, –. Leiden: Brill. Michaelis, Laura A. (). Aspect and semantics–pragmatics interface: The case of already. Lingua : –. Michaelis, Laura A. (). On the use and meaning of already. Linguistics and Philosophy : –. Miestamo, Matti (). Clausal Negation: A Typological Study. Helsinki: University of Helsinki Ph.D. dissertation. Miestamo, Matti (). Standard Negation: The Negation of Declarative Verbal Main Clauses in a Typological Perspective. Berlin: Mouton de Gruyter. Miestamo, Matti (). Negatives without negators. In Jan Wohlgemuth and Michael Cysouw (eds.), Rethinking Universals: How Rarities Affect Linguistic Theory, –. Berlin: de Gruyter Mouton. Miestamo, Matti (). A typological perspective on negation in Finnish dialects. Nordic Journal of Linguistics : –. Miestamo, Matti (a). Symmetric and asymmetric standard negation. In Dryer and Haspelmath (eds.) , accessed February , . Miestamo, Matti (b). Subtypes of asymmetric standard negation. In Dryer and Haspelmath (eds.) , accessed September , .

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Miestamo, Matti (). Negation. In Alexandra Y. Aikhenvald and R. M. W. Dixon (eds.), Cambridge Handbook of Linguistic Typology, –. Cambridge: Cambridge University Press. Miestamo, Matti, and Johan van der Auwera (). Negative declaratives and negative imperatives— similarities and differences. In Andreas Ammann (ed.), Linguistics Festival. May , Bremen, –. Bremen: Brockmeyer. Miestamo, Matti, Dik Bakker, and Antti Arppe (). Sampling for variety. Linguistic Typology : –. Miestamo, Matti, Kaius Sinnemäki, and Fred Karlsson (eds.) (). Language Complexity: Typology, Contact, Change. Amsterdam: Benjamins. Miller, Karen (). Sociolinguistic variation in Brown’s Sarah corpus. In Alia K. Biller, Esther Y. Chung, and Amelia E. Kimball (eds.), Proceedings of the th Annual Boston University Conference on Child Language Development (BUCLD) : –. Milner, Jean-Claude (). De la syntaxe à l’interprétation: Quantités, insultes, exclamations. Paris: Éditions du Seuil. Milner, Jean-Claude (). Le système de la négation en français et l’opacité du sujet. Langue française : –. Milroy, James (). Language ideologies and the consequences of standardization. Journal of Sociolinguistics  (): –. Mixco, Mauricio J. (). Mandan. Munich: Lincom Europa. Miyake, Yoshimi (). Reduplication in Javanese. Asian and African Languages and Linguistics . Mody, S., and S. Carey (). The emergence of reasoning by the disjunctive syllogism in early childhood. Cognition : –. Mody, S., R. Feiman, and S. Carey (submitted). Domain-general negation in infants’ reasoning. Moeschler, Jacques (). Why are there no negative particulars? Horn’s conjecture revisited. GG@G [Generative Grammar at Geneva] : –. Moeschler, Jacques (). Negation, scope and the descriptive/metalinguistic distinction. Generative Grammar in Geneva : –. Moeschler, Jacques (). How “Logical” are Logical Words? Negation and its Descriptive vs. Metalinguistic Uses. In Maite Taboada and Radoslava Trnavac (eds.), Nonveridicality and Evaluation: Theoretical, Computational and Corpus Approaches, –. Leiden: Brill. Moeschler, Jacques (). Qu’y a-t-il de représentationnel dans la négation métalinguistique? Nouveaux cahiers de linguistique française : –. Moeschler, Jacques (). Where is procedural meaning? Evidence from discourse connectives and tenses. Lingua –: –. Moeschler, Jacques (a). Back to negative particulars: A truth-conditional account. In Stavros Assimakopoulos (ed.), Pragmatics at its Interfaces, –. Berlin: Mouton de Gruyter. Moeschler, Jacques (b). Formal and natural languages: What does logic tell us about natural language? In Anne Barron, Yueguo Gu, and Gerard Steen (eds.), The Routledge Handbook of Pragmatics, –. London: Routledge. Moeschler, Jacques (c). Representation and metarepresentation in negation. In R. Carston, K. Scott, and B. Clark (eds.), Relevance Theory: Pragmatics and Interpretation. Cambridge: Cambridge University Press. Moeschler, Jacques (). A set of semantic and pragmatic criteria for descriptive vs. metalinguistic negation. Glossa: A Journal of General Linguistics  (): .–. Moeschler, Jacques (). Non-Lexical Pragmatics: Time, Causality and Logical Words. Berlin: Mouton de Gruyter. Mohamed, Mona A., David M. Yousem, Aylin Tekes, Nina Browner, and Vince D. Calhoun (). Correlation between the amplitude of cortical activation and reaction time: A functional MRI study. American Journal of Roentgenology  (): –. Moll, Henrike, and Michael Tomasello (). Cooperation and human cognition: The Vygotskian intelligence hypothesis. Phil. Trans. R. Soc. B : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Montague, Richard (). The proper treatment of quantification in ordinary English. In K. J. J. Hintikka, J. M. E. Moravcsik, and P. Suppes (eds.), Approaches to Natural Language, –. Dordrecht: Reidel. Montébran, Aurore (). Negation in the context of TAM: The example of Hula, Mankanya. Presentation to the workshop Negation and Negativity in Natural Language. Ruhr Universität Bochum, April . Monti Martin, M., Larry Parsons, M. J. Martinez, and Daniel N. Osherson (). Functional neuroanatomy of deductive inference: A language-independent distributed network. NeuroImage : –. Montrul, Silvina (). The Acquisition of Heritage Languages. Cambridge: Cambridge University Press. Moodie, Jonathan (). A Grammar of the Lopit Language. Ph.D. dissertation. University of Melbourne. Morgan, Jerry L. (). Sentence fragments and the notion “sentence”. In Issues in Linguistics: Papers in Honor of Henry and Renée Kahane. Urbana, IL: University of Illinois Press. Morgan, Jerry L. (). Two types of convention in indirect speech acts. In Peter Cole (ed.), Syntax and Semantics : Pragmatics, –. New York: Academic Press. Morgenstern, Aliyah, Pauline Beaupoil Hourdel, Marion Blondel, and Dominique Boutet (). A multimodal approach to the development of negation in signed and spoken languages: Four case studies. In L. Ortega, A. E. Tyler, H. I. Park, and M. Uno (eds.), The UsageBased Study of Language Learning and Multilingualism, –. Washington, DC: Georgetown University Press. Morton, Deborah (). The temporal and aspectual semantics and verbal tonology of Gisida Anii. Ph.D. dissertation, The Ohio State University. Moscati, Vincenzo (). The Scope of Negation. Siena: University of Siena dissertation. Moscati, Vincenzo (). Negation Raising: Logical Form and Linguistic Variation. Newcastle: Cambridge Scholars Publishing. Mosegaard-Hansen, Maj-Britt (). Negation in the history of French. In David Willis, Christopher Lucas, and Anne Breitbarth (eds.), The History of Negation in the Languages of Europe and the Mediterranean, Vol. : Case Studies, –. Oxford: Oxford University Press. Mosonyi, Esteban Emilio, Jorge Carlos Mosonyi, and Pedro J. Medina Tamanaico (). Kariña [Kari’ña], in Esteban Emilio Mosonyi and Jorge Carlos Mosonyi (eds.), Manuel de lenguas indígenas de Venezuela, Vol , –. Caracas: Fundación Bigott. Moxey, L. M., and A. J. Sanford (). Quantifiers and focus. Journal of Semantics : –. Moxey, L. M., A. J. Sanford, and E. J. Dawydiak (). Denials as controllers of negative quantifier focus. Journal of Memory and Language  (): –. Moyse-Faurie, Claire (). Grammaire du futunien. Noumea: Centre de Documentation Pédagogique. Moyse-Faurie, Claire (). Negation in East Futunan (Futuna, Wallis and Futuna Islands). In Hovdhaugen and Mosel (eds.), Negation in Oceanic Languages, –. Mughazy, Mustafa (). Metalinguistic negation and truth functions: The case of Egyptian Arabic. Journal of Pragmatics : –. Muller, Claude (). La négation explétive dans les constructions complétives. Langue française : –. Müller, Natascha, and Aafke Hulk (). Crosslinguistic influence in bilingual language acquisition: Italian and French as recipient languages. Bilingualism: Language and Cognition. (): –. Murphy, M. Lynne. (). Semantic Relations and the Lexicon. Cambridge: Cambridge University Press. Murzi, Julien, and Ole Thomassen Hjortland (). Inferentialism and the categoricity problem: Reply to Raatikainen. Analysis  (): –. Musolino, J., and J. Lidz (). Why children aren’t universally successful with quantification. Linguistics : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Mutu, Margaret (with Ben Teìkitutoua) (). Ùa Pou: Aspects of a Marquesan Dialect. Canberra: Pacific Linguistics. Muysken, Pieter, Harald Hammarström, Joshua Birchall, Swintha Danielsen, Love Eriksen, Ana Vilacy Galucio, Rik van Gijn, Simon van de Kerke, Vishnupraya Kolipakam, Olga Krasnoukhova, Neele Müller, and Loretta O’Connor (). The languages of South America: Deep families, areal relationships, and language contact. In Loretta O’Connor and Pieter Muysken (eds.), The Native Languages of South America: Origins, Development, Typology, –. Cambridge: Cambridge University Press. Nakanishi, Kimiko (). The scope of even and quantifier raising. Natural Language Semantics : –. Nakata, H., K. Sakamoto, A. Ferretti, M. Gianni Perrucci, C. Del Gratta, R. Kakigi, and G. Luca Romani (). Somato-motor inhibitory processing in humans: An event-related functional. MRI Study Neuroimage : –. Napoli, D. J., and M. Nespor (). Negatives in comparatives. Language  (): –. Nastase, Samuel A., and James V. Haxby (). Structural basis of semantic memory. In Learning and Memory: A Comprehensive Reference (nd edn.), – . Natsopoulos, Dimitris (). A verbal illusion in two languages. Journal of Psycholinguistic Research  (): –. Nawroth, Christian, and Eberhard von Borell (). Domestic pigs’ (Sus scrofa domestica) use of direct and indirect visual and auditory cues in an object choice task. Animal Cognition : –. Neale, Stephen (). The philosophical significance of Gödel’s Slingshot. Mind  (): –. Neale, Stephen (). Facing Facts. Oxford: Clarendon. Neale, Stephen, and Josh Dever (). Slingshots and boomerangs. Mind  (): –. Nedwick, Kelly M. (). Metalinguistic Negation in English and Arabic. Ph.D. dissertation. University of Yale. Nedyalkov, Igor (). Evenki. In Peter Kahrel and René van den Berg (eds.), Typological Studies in Negation, –. Amsterdam: Benjamins. Neeleman, Ad, and Hans van de Koot (). On syntactic and phonological representations. Lingua  (): –. Neeleman, Ad, and Hans van de Koot (). Dutch scrambling and the nature of discourse templates. The Journal of Comparative Germanic Linguistics  (): –. Neeleman, Ad, and Reiko Vermeulen (eds.) (). The Syntax of Topic, Focus, and Contrast: An Interface-Based Approach. Vol. . Berlin: Walter de Gruyter. Neidle, Carol, Judy Kegl, Dawn MacLaughlin, Ben Bahan, and Robert G. Lee (). The Syntax of American Sign Language: Functional Categories and Hierarchical Structure. Cambridge, MA: MIT Press. Nevalainen, Terttu (). Negative concord as an English “Vernacular Universal”: Social history and linguistic typology. Journal of English Linguistics : –. Newmeyer, Frederick J. (). Peut-on reconstruire la langue des premiers humains? Nouveaux cahiers de linguistique française : –. Ng, Anne, and E. Matthew Husband (). Interference across the at-issue/not-at-issue divide: Agreement and NPI licensing. Poster at th Annual CUNY Conference on Human Sentence Processing, Massachusetts Institute of Technology. Ngangoum, Emilienne (). When Synchrony Meets Diachrony: The Negation System of FeɁfεɁ in the Light of the Jespersen Cycle. Doctoral dissertation Utrecht University. Ngata, H. M. (). English–Maori Dictionary. Wellington: Learning Media. Nicol, J., and D. Swinney (). The role of structure in coreference assignment during sentence comprehension. Journal of Psycholinguistic Research  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Nicolae, Andreea Cristina (). Positive polarity items: An alternative-based account. In Ana Aguilar Guevara, Anna Chernilovskaya, and Rick Nouwen (eds.), Proceedings of Sinn und Bedeutung, : –. Nicolae, Andreea Cristina (). Any questions? Polarity as a window into the structure of questions. Ph.D. dissertation, Harvard University. Nieuwland, M. S. (). Quantification, prediction, and the online impact of sentence truth-value: Evidence from event-related potentials. Journal of Experimental Psychology: Learning, Memory, and Cognition : –. Nieuwland, M. S., and G. R. Kuperberg (). When the truth is not too hard to handle: An event-related potential study on the pragmatics of negation. Psychological Science  (): –. Nieuwland, M. S. and A. E. Martin (). If the real world were irrelevant, so to speak: The role of propositional truth-value in counterfactual sentence comprehension. Cognition : –. Nigbur, R., G. Ivanova, and B. Stürmer (). Theta power as a marker of cognitive interference. Clin Neurophysiol : –. Nilsen, Øystein (). Domains for adverbs. Lingua : –. Noh, Eun-Ju (). Echoquestions: Metarepresentation and pragmatic enrichment. Linguistics and Philosophy  (): –. Noh, Eun-Ju (). Metarepresentation: A Relevance-Theory Approach. Amsterdam: John Benjamins Publishing. Noh, Eun-Ju, Hyeree Choo, and Sungryong Koh (). Processing metalinguistic negation: Evidence from eye-tracking experiments. Journal of Pragmatics : –. Nordmeyer, Ann, and Michael Frank (unpublished manuscript). Negation is only hard to process when it is pragmatically infelicitous. . Nordmeyer, Ann, and Michael Frank (). The role of context in young children’s comprehension of negation. Journal of Memory and Language : –. Nordmeyer, Ann, and Michael Frank (). Negation is only hard to process when it is pragmatically infelicitous. Proceedings of the th Annual Meeting of the Cognitive Science Society, –. Nordmeyer, Ann, and Michael Frank (). Early understanding of pragmatic principles in children’s judgements of negative sentences. Language Learning and Development : –. Northrup, Oliver Burton (). Grounds for Commitments. Ph.D. dissertation, University of California Santa Cruz. Nouwen, Rick (). Upper-bound no more: The exhaustive interpretation of nonstrict comparison. Natural Language Semantics : –. Nunes, Jairo, and Ronice Müller de Quadros (). Phonetically realized traces in American Sign Language and Brazilian Sign Language. In Josep Quer (ed.), Signs of the Time: Selected Papers from TISLR , –. Hamburg: Signum Press. Nyvad, Anne Mette, Ken Ramshøj Christensen, and Sten Vikner (). The left periphery and agrammatism. In Enoch Oladé Aboh, Maria Teresa Guasti, and Ian Roberts (eds.), Locality, –. Oxford: Oxford University Press. Obenauer, Hans-Georg (). Special interrogatives—Left periphery, wh-doubling and (apparently) optional elements. In Jenny Doetjes and Paz González (eds.), Romance Languages and Linguistics Theory , –. Amsterdam/Philadelphia: Benjamins. Obiamalu, Greg Orji (). On the role of tone in Igbo negation. Journal of West African Languages  (): –. Obiamalu, Greg Orji (). Negation and V-movement in Igbo: A minimalist perspective. SKASE Journal of Theoretical Linguistics  (): –. Ogihara, Toshiyuki (). The semantics of tense in embedded clauses. Linguistic Inquiry  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





O’Grady, William (). Does emergentism have a chance? In H. Chan, H. Jacob, and E. Kapia (eds.), Proceedings of the nd Annual Boston University Conference on Language Development (BUCLD), –. Somerville, MA: Cascadilla Press. O’Leary, Carrie, and Stephen Crain (). Negative polarity (a positive result) and positive polarity (a negative result). Paper presented at the th Annual Boston University Conference on Child Language Development, Boston. Olza Moreno, Inés (). Metalinguistic negation and explicit echo, with reference to English and Spanish. In Malin Roitman (ed.), The Pragmatics of Negation: Negative Meanings, Uses and Discursive Functions, –. Amsterdam/Philadelphia: John Benjamins. Oomen, Marloes, and Roland Pfau (). Signing NOT (or not): A typological perspective on standard negation in Sign Language of the Netherlands. Linguistic Typology  (): –. Oomen, Marloes, Roland Pfau, and Enoch O. Aboh (). High and low negation in Sign Language of the Netherlands. FEAST : –. Orenes, I., and C. Santamaría (). Visual content of words delays negation. Acta psychologica : –. Orenes, I., D. Beltrán, and C. Santamaría (). How negation is understood: Evidence from the visual world paradigm. Journal of Memory and Language : –. Orenes, I., L. Moxey, C. Scheepers, and C. Santamaría (). Negation in context: Evidence from the visual world paradigm. Quarterly Journal of Experimental Psychology : –. Orlandini, Anna (). Négation et argumentation en latin. Grammaire fondamentale du latin, Tome VIII. Louvain: Peeters. Ortega, Lourdes (). Trying out theories on interlanguage description and explanation over  years of L negation research. In Z. Han and E. Tarone (eds.), Interlanguage:  Years Later, –. Amsterdam: John Benjamins. Ortiz de Urbina, Jon (). Parameters in the Grammar of Basque. Dordrecht: Foris. Oseki, Yohei (). The syntax of denial. Unpublished paper presented at the workshop on New Perspectives of Generative Grammar, Hokkaido University, November , , and at ELSJ th International Spring Forum, April –, . O’Shaughnessy, Douglas, and Joseph Allen (). Linguistic modality effects on fundamental frequency in speech. Journal of the Acoustical Society of America  (): –. Osterhout, Lee, and Phillip Holcomb (). Event-related brain potentials elicited by syntactic anomaly. J. Mem. Lang. : –. Osterhout, Lee, and Janet Nicol (). On the distinctiveness, independence, and time course of the brain responses to syntactic and semantic anomalies. Language and Cognitive Processes  (): –. Osterhout, Lee, Phillip J. Holcomb, and David A. Swinney (). Brain potentials elicited by gardenpath sentences: Evidence of the application of verb information during parsing. Journal of Experimental Psychology: Learning, Memory, and Cognition  (): –. Otani, K., and J. Whitman (). V-Raising and VP-Ellipsi. Linguistic Inquiry : –. Otsuka, Yuko (). Split ergativity and the nature of pronouns in Tongan. In Andrea Rackowski and Norvin Richards (eds.), Proceedings of AFLA : The Eighth Meeting of the Austronesian Formal Linguistics Association, –, MIT Working Papers in Linguistics . Cambridge, MA: MITWPL. Ott, Dennis, and Volker Struckmeier (). Particles and deletion. Linguistic Inquiry  (): –. Ouali, Hamid (). Sentential negation in Berber: A comparative study. In John Mugany (ed.), Linguistic Description: Typology and Representation of African Languages, –. Trends in African Linguistics . Trenton, NJ: African World Press. Ouali, Hamid (). On C-to-T ϕ-feature transfer: The nature of agreement and anti- agreement in Berber. In Roberta D’Alessandro, Gunnar Hrafnbjargarson, and Susann Fischer (eds.), Agreement Restrictions, –. Berlin: Mouton de Gruyter.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Ouali, Hamid (). Agreement, Pronominal Clitics and Negation in Tamazight Berber: A Unified Analysis. London: Continuum. Ouhalla, Jamal (). The Syntax of Head Movement: A Study of Berber. Ph.D. dissertation, University College London. Ouhalla, Jamal (). Sentential negation, Relativized Minimality and the aspectual status of auxiliaries. Linguistic Review : –. Owens, Jonathan (). A Grammar of Harrar Oromo (Northeastern Ethiopia). Hamburg: Helmut Buske. Pagliarini, Elena, Stephen Crain, and Maria Teresa Guasti (). The compositionality of logical connectives in child Italian. Journal of Psycholinguistic Research : –. Pakulak, E., and H. Neville (). Proficiency differences in syntactic processing of monolingual native speakers indexed by event-related potentials. Journal of Cognitive Neuroscience : –. Palacios Martínez, Ignacio (). Negative concord in the language of British adults and teenagers. English World-Wide: A Journal of Varieties of English  (): –. Pallier, C., A.-D. Devauchelle, and S. Dehaene (). Cortical representation of the constituent structure of sentences. Proceedings of the National Academy of Sciences  (): –. Papeo, Liuba, and Alfonso Caramazza (). When “ultrarapid” word-related motor activity is not faster than “early”. Frontiers in Human Neuroscience : . Papeo, Liuba, and Jean-Rémy Hochmann (). A cross-talk between brain-damage patients and infants on action and language. Neuropsychologia : –. Papeo, Liuba, Jean-Rémy Hochmann, and Lorella Battelli (). The default computation of negated meanings. Journal of Cognitive Neuroscience : –. Papeo, Liuba, Avaro Pascual-Leone, and Alfonso Caramazza (). Disrupting the brain to validate hypotheses on the neurobiology of language. Frontiers in Human Neuroscience : . Papeo, Liuba, Antonino Vallesi, Alessio Isaja, and Raffaella I. Rumiati (). Effects of TMS on different stages of motor and non-motor verb processing in the primary motor cortex. PLoS ONE (): e. Papeo, Liuba, Raffaella I. Rumiati, Cinzia Cecchetto, and Barbara Tomasino (). On-line changing of thinking about words: The effect of cognitive context on neural responses to verb reading. Journal of Cognitive Neuroscience : –. Papeo, Liuba, Angelika Lingnau, Sara Agosta, Alvaro Pascual-Leone, Lorella Battelli, and Alfonso Caramazza (). The origin of word-related motor activity. Cerebral Cortex : –. Paradis, Carita, Caroline Willners, and Steven Jones (). Good and bad opposites: Using textual and experimental techniques to measure antonym canonicity. The Mental Lexicon : –. Paradis, Johanne, and Fred Genesee (). On continuity and the emergence of functional categories in bilingual first language acquisition. Language Acquisition : –. Park, Tschang-Zin () Some facts on negation: Wode’s four-stage developmental theory of negation revisited. J Child Lang : –. Parker, D. and C. Phillips (). Illusory negative polarity item licensing is selective. Poster at the th CUNY Conference on Human Sentence Processing. Stanford University. Parker, D., and C. Phillips (). Negative polarity illusions and the format of hierarchical encodings in memory. Cognition : –. Parker, D., and C. Phillips (). Reflexive attraction in comprehension is selective. Journal of Memory and Language : –. Parsons, Terence (). Assertion, denial, and the liar paradox. Journal of Philosophical Logic : –. Parsons, Terence (). Events in the Semantics of English: A Study in Subatomic Semantics. Cambridge, MA: MIT Press. Parsons, Terence (). The traditional Square of Opposition. In E. Zalta (ed.), Stanford Encyclopedia of Philosophy. .

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Partee, Barbara H. (). Some structural analogies between tenses and pronouns in English. Journal of Philosophy  (): –. Partee, Barbara H. (). On the “scope of negation” and polarity sensitivity. In Eva Hajičová (ed.), Functional Description of Language, –. Prague: Faculty of Mathematics and Physics, Charles University. Paterson, K. B., R. Filik, and S. P. Liversedge (). Competition during the processing of quantifier scope ambiguities: Evidence from eye movements during reading. Quarterly Journal of Experimental Psychology  (): –. Paterson, K. B., A. J. Sanford, L. M. Moxey, and E. Dawydiak (). Quantifier polarity and referential focus during reading. Journal of Memory and Language : –. Payne, John R. (). Negation. In Timothy Shopen (ed.), Language Typology and Syntactic Description, Vol. : Clause Structure, –. Cambridge: Cambridge University Press. Pea, Roy D. (a). The development of negation in early child language. In D. Olson (ed.), The Social Foundations of Language and Thought, –. New York: W. W. Norton. Pea, Roy D. (b). Logic in Early Child Language. Annals of the New York Academy of Sciences : –. Pearce, Elizabeth (). Negation and indefinites in Maori. In Danielle Forget, Paul Hirschbühler, France Martineau, and María Luisa Rivero (eds.), Negation and Polarity: Syntax and Semantics, –. Amsterdam: Benjamins. Pearce, Elizabeth (). Topic and focus in a head-initial language: Māori. In Carolyn Smallwood and Katherine Kitto (eds.), Proceedings of AFLA VI: The Sixth Meeting of the Austronesian Formal Linguistics Association, –. Toronto: University of Toronto. Pearce, Elizabeth (). Preverbal subjects and preverbal particles in Māori. Paper presented at the Workshop on the Syntax of Polynesian Languages, University of Toronto, June –. Peeke, Catherine (). Waorani. In Kahrel and van den Berg (eds.), Typological Studies in Negation, –. Penello, Nicoletta, and Diego Pescarini (). Osservazioni su mica in italiano e alcuni dialetti veneti. In Federica Cognola and Diego Pescarini (eds.), La negazione: Variazione dialettale ed evoluzione diacronica, –. Quaderni di lavoro ASIt . Padua: Unipress. Penka, Doris (). Negative Indefinites. Oxford: Oxford University Press. Penka, Doris, and Arnim von Stechow (). Negative Indefinita unter Modalverben. In Reimar Müller and Marga Reis (eds.), Modalität und Modalverben im Deutschen, –. Hamburg: Helmut Buske Verlag. Pepperberg, Irene M., Mark R. Willner, and Lauren B. Gravitz (). Development of Piagetian object permanence in grey parrot (Psittacus erithacus). Journal of Comparative Psychology : –. Perales, Susana (). The status of the auxiliary do in L and L English negative clauses. IRAL, International Review of Applied Linguistics in Language Teaching : –. Pereltsvaig, Asya (). Negative polarity items in Russian and the “Bagel problem”. In Adam Przepiorkowski and Sue Brown (eds.), Negation in Slavic, –. Bloomington, IL: Slavica Publishers. Pesetsky, David (). Wh-in-situ: Movement and unselective binding. In Eric J. Reuland and Alice G. B. ter Meulen (eds.), The Representation of (In)definiteness, –. Cambridge, MA: MIT Press. Pesetsky, David (). Phrasal Movement and its Kin. Cambridge, MA: MIT Press. Petronio, Karen (). Clause Structure in American Sign Language. Ph.D. dissertation, University of Washington. Petronio, Karen, and Diane Lillo-Martin (). WH-movement and the position of Spec-CP: Evidence from American Sign Language. Language  (): –. Pfau, Roland (). The grammar of headshake: A typological perspective on German Sign Language negation. Linguistics in Amsterdam  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Pfau, Roland (). The grammaticalization of head shake: From head movement to negative head. In Andrew D. M. Smith, Graeme Trousdale, and Richard Waltereit (eds.), New Directions in Grammaticalization Research, –. Amsterdam: John Benjamins Publishing Company. Pfau, Roland (). A featural approach to sign language negation. In Pierre Larrivée and Chungmin Lee (eds.), Negation and Negative Polarity: Cognitive and Experimental Perspectives, –. Dordrecht: Springer. Pfau, Roland, and Josep Quer (). V-to-Neg Raising and negative concord in three sign languages. Rivista di Grammatica Generativa : –. Pfau, Roland, and Josep Quer (). On the syntax of negation and modals in German Sign Language (DGS) and Catalan Sign Language (LSC). In Pamela Perniss, Roland Pfau, and Markus Steinbach (eds.), Visible Variation: Cross-linguistic Studies in Sign Language Structure, –. Berlin: Mouton de Gruyter. Pfau, Roland, and Josep Quer (). Nonmanuals: Their grammatical and prosodic roles. In Diane Brentari (ed.), Cambridge Language Surveys: Sign Languages, –. Cambridge: Cambridge University Press. Pfau, Roland, and Markus Steinbach (). Modality-independent and modality-specific aspects of grammaticalization in sign languages. Linguistics in Potsdam : – (available at ) Pfau, Roland, and Markus Steinbach (). Headshakes in Jespersen’s Cycle. Paper presented at th Conference on Theoretical Issues in Sign Language Research (TISLR ), London, July . Pfurtscheller, Gert, and F. H. Lopes Da Silva (). Event-related EEG/MEG synchronization and desynchronization: Basic principles. Clinical Neurophysiology : –. Pheby, John (). Intonation und Grammatik im Deutschen. Berlin: Akademie-Verlag. Piaget, Jean (). The Construction of Reality in the Child. New York: Basic Books. Picallo, M. Carme (). Modal verbs in Catalan. Natural Language and Linguistic Theory  (): –. Pierrehumbert, Janet B., and Julia Hirschberg (). The meaning of intonational contours in the interpretation of discourse. In Philip R. Cohen, Jerry Morgan, and Martha E. Pollack (eds.), Intentions in Communication, –. Cambridge, MA: MIT Press. Pietarinen, Ahti-Veikko (). What is a negative polarity item? Linguistic Analysis  (–): –. Pietroski, Paul (). Causing Actions. Oxford: Oxford University Press. Pietroski, Paul (). Events and Semantic Architecture. Oxford: Oxford University Press. Pietroski, Paul (). Framing event variables. Erkenntnis  (): –. Pinto, Clara (). Negação metalinguística e estruturas com nada no português europeu. MA dissertation. University of Lisbon. Pinto, Clara (). Negative indefinites in the history of Portuguese: The case of nemigalha. Estudos de lingüística Galega : –. Pintzuk, Susan (). Phrase Structures in Competition: Variation and Change in Old English Word Order. Doctoral Dissertation, University of Pennsylvania. Pitts, Alyson (). Exploiting a “pragmatic ambiguity” of negation. Language  (): –. Plumwood, Val (). The politics of reason: Towards a feminist logic. Australasian Journal of Philosophy  (): –. Poletto, Cecilia (). On negative doubling. In Federica Cognola and Diego Pescarini (eds.), La negazione: Variazione dialettale ed evoluzione diacronica, –. Quaderni di lavoro ASIt . Padua: Unipress. Poletto, Cecilia (). The syntax of focus negation. Quaderni di lavoro dell’ASIt : –. Poletto, Cecilia (). Comparative linguistics and microvariation: The role of Dialectology. Language in Context  (): –. Poletto, Cecilia (). Word Order in Old Italian. Oxford: Oxford University Press.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Poletto, Cecilia (). Negative doubling: In favor of a big NegP analysis. In Silvio Cruschina, Katharina Hartmann, and Eva-Maria Remberger (eds.), Studies on Negation: Syntax, Semantics, and Variation, –. Göttingen: Vienna University Press. Poletto, Cecilia (). L’etimologia come finestra sulla sintassi: Evidenza empirica dai dialetti italiani. In M. Glessgen et al. (eds.), Repenser la variation linguistique, –. Strasbourg: Éditions de linguistique et de philologie. Poletto, Cecilia, and Raffaella Zanuttini (). Emphasis as reduplication : Evidence from si che/no che sentences. Lingua : –. Polinsky, Maria (). Deconstructing Ergativity: Two Types of Ergative Languages and their Features. New York: Oxford University Press. Pollock, Jean-Yves (). Verb movement, Universal Grammar and the structure of IP. Linguistic Inquiry  (): –. Poole, Geoffrey (). Focus and the development of N-words in Spanish. In Janine Berns, Haike Jacobs, and Tobias Scheer (eds.), Romance Languages and Linguistic Theory : Selected Papers from “Going Romance” Nice , –. Amsterdam: Benjamins. Pope, Emily (). Questions and Answers in English. The Hague: Mouton. Porcellato, Damiana (). La negazione standard: Uno studio tipologico sui marcatori di negazione. Master Thesis, University of Padua. Posner, Rebecca (). Double negatives, negative polarity and negative incorporation in Romance: A historical and comparative view. Transactions of the Philological Society : –. Posner, Rebecca (). Post-verbal negation in non-standard French: A historical and comparative view. Romance Philology  (): –. Postal, Paul M. (a). Skeptical Linguistic Essays. Oxford: Oxford University Press. Postal, Paul M. (b). The structure of one type of American English vulgar minimizer. In Paul Postal (ed.), Skeptical Linguistic Essays, –. Oxford/New York: Oxford University Press. Potsdam, Eric (). CP-negation and the domain of quantifier raising. Linguistic Inquiry  (): –. Pott, August F. (). Etymologische Forschungen auf dem Gebiete der Indo-Germanischen Sprachen, Vol. . Lemgo/Detmold: Meyer. Potts, Chris (). The Logic of Conventional Implicatures. Oxford: Oxford University Press. Potts, Chris (). Presupposition and implicature. In S. Lappin and C. Fox (eds.), The Handbook of Contemporary Semantic Theory, –. New York: Wiley. Poulos, George and Sonja E. Bosch (). Zulu. Munich: Lincom Europa. Poutsma, Hendrik (). A Grammar of Late Modern English. Groningen: P. Noordhoff. Pratt, N., and H. A. Whitaker (). Aphasia syndromes. In Keith Brown (ed.), Encyclopedia of Language & Linguistics (nd edn.), –. Oxford: Elsevier. Prawitz, Dag (). Meaning and proofs: On the conflict between classical and intuitionistic logic. Theoria  (): –. Prévost, Phillipe, and Lydia White (). Missing surface inflection or impairment in second language acquisition? Evidence from tense and agreement. Second Language Research  (): –. Price, Huw (). Sense, assertion, Dummett, and denial. Mind  (): –. Price, Huw (). Why “not”? Mind  (): –. Price, Huw (n.d.). ‘Not’ again. Priest, Graham (). Towards Non-Being: The Logic and Metaphysics of Intentionality. Oxford: Oxford University Press. Priest, Graham (). Doubt Truth to Be a Liar. Oxford: Oxford University Press. Prieto, Pilar (). Intonational meaning. WIRES Cognitive Science : –. Prieto, Pilar, Joan Borràs-Comes, Susagna Tubau, and M.Teresa Espinal (). Prosody and gesture constrain the interpretation of double negation. Lingua : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Prince, Ellen F. (). The syntax and semantics of neg-raising, with evidence from French. Language : –. Prince, Ellen F. (). A comparison of WH-Clefts and IT-Clefts in discourse. Language : –. Prince, Ellen F. (). Toward a taxonomy of given/new information. In Peter Cole (ed.), Radical Pragmatics, –. New York: Academic Press. Prince, Ellen F. (). On the syntactic marking of presupposed open propositions. In A. Farley, P. Farley, and K.-E. McCullough (eds.), Papers from the Parasession on Pragmatics and Grammatical Theory, nd Regional Meeting, Chicago Linguistic Society, –. Prior, A. N. (). Negation. Encyclopedia of Philosophy : –. New York: Macmillan. Prior, A. N. () Negation. In D. M. Borchert (ed.), Encyclopedia of Philosophy, nd edn., Vol. , –. New York: Macmillan. Progovac, Ljiljana (). A Binding Approach to Polarity Sensitivity. Ph.D. thesis, University of Southern California. Progovac, Ljiljana (). Negative polarity: A semantico-syntactic approach. Lingua : –. Progovac, Ljiljana (). Negative polarity: Entailment and binding. Linguistics and Philosophy  (): –. Progovac, Ljiljana (). Negative and Positive Polarity: A Binding Approach. Cambridge: Cambridge University Press. Progovac, Ljiljana (). Negative and positive feature checking and the distribution of polarity items. In Sue Brown and Adam Przepiórkowski (eds.), Negation in Slavic, –. Bloomington, IN: Slavica Publishers. Progovac, Ljiljana, Kate Paesani, Eugenia Casielles, and Ellen Barton (eds.) (). The Syntax of Nonsententials: Multidisciplinary Perspectives (Linguistik Aktuell/Linguistics Today ). Amsterdam/New York: John Benjamins. Przepiórkowski, Adam, and Anna Kupc (). Eventuality negation and negative concord in Polish and Italian. In Robert D. Borsley and Adam Przepiórkowski (eds.), Slavic in Head-Driven Phrase Structure Grammar, –. Stanford, CA: CSLI Publications. Pulvermüller, Friedemann (). Brain reflections of words and their meaning. Trends in Cognitive Sciences  (): –. Pulvermüller, Friedemann (). Brain mechanisms linking language and action. Nature Reviews Neuroscience : –. Puskás, Genoveva (). Licensing double negation in NC and non-NC languages. Natural Language and Linguistic Theory : –. Putnam, Michael, and Liliana Sánchez (). What’s so incomplete about incomplete acquisition? A prolegomenon to modeling heritage language grammars. Linguistic Approaches to Bilingualism (): –. Quadros, Ronice M. de (). Phrase Structure of Brazilian Sign Language. Ph.D. dissertation, Pontifíca Universidade Católica do Rio Grande do Sul, Porto Alegre. Quer, Josep (). The Licensing of Negative Items. MA thesis, Autonomous University of Barcelona. Quer, Josep (). Interpreting mood. Probus : –. Quer, Josep (/). Operadores negativos en Lengua de Signos Catalana. In Sandra B. Cvejanov (ed.), Lenguas de señas: Estudios de lingüística teórica y aplicada, –. Neuquén: Editorial de la Universidad Nacional del Comahue, Argentina. Quer, Josep (). Negation. In Roland Pfau, Markus Steinbach, and Bencie Woll (eds.), Sign Language: An International Handbook (Handbooks of Linguistics and Communication Science, HSK), –. Berlin: Mouton de Gruyter. Quer, Josep, and Rosa Ma Boldú (). Lexical and morphological resources in the expression of sentential negation in Catalan Sign Language (LSC). Actes del è Congrés de Lingüística General, Universitat de Barcelona. CD-ROM.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Quine, Willard Van Orman (). Semantics and abstract objects. In Proceedings of the American Academy of Arts and Sciences  (): –. Quine, Willard Van Orman (). Word and Object. Cambridge, MA: MIT Press. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik (). A Comprehensive Grammar of the English Language. London and New York: Longman. Raichle, Marcus E. (). Two views of brain function. Trends in Cognitive Sciences  (): –. Raichle, Marcus E., Ann Mary MacLeod, Abraham Z. Snyder, William J. Powers, Debra A. Gusnard, and Gordon L. Shulman (). A default mode of brain function. Proceedings of the National Academy of Sciences  (): –. Ramat, Paolo (). Perché veruno significa nessuno? In Paolo Ramat and Elisa Roma (eds.), Sintassi storica. Atti del XXX Congresso Internazionale della Società di Linguistica Italiana, –. Rome: Bulzoni. Randal, Scott (). A grammatical sketch of Tennet. In Gerrit Jan Dimmendaal and Marco Last (eds.), Surmic Languages and Cultures, –. Cologne: Rüdiger Köppe Verlag. Rankin, Tom (). The transfer of V: Inversion and negation in German and Dutch learners of English. International Journal of Bilingualism  (): –. Rausch, P. J. (). Die Sprache von Südost-Bougainville, Deutsche Salomonsinseln. Ahtropos : –, –, –. R. Core Team (). R: A Language and Environment for Statistical Computing. Vienna: R Foundation for Statistical Computing. Reboul, Anne (). Causalité, force dynamique et ramifications temporelles. Cahiers de linguistique française : –. Reese, Brian J. (). The meaning and use of negative polar interrogatives. In O. Bonami and P. Cabredo-Hofherr (eds.), Empirical Issues in Syntax and Semantics, : –. . Reese, Brian J. (). Bias in Questions. Ph.D. dissertation, University of Texas at Austin. Reesink, Ger P. (a). The eastern Bird’s Head languages compared. In Ger P. Reesink (ed.), Languages of the Eastern Bird’s Head, –. Canberra: The Australian National University. Reesink, Ger P. (b). Clause-final negation, structure and interpretation. Functions of Language : –. Reich, Ingo (). Toward a uniform analysis of short answers and gapping. In Kerstin Schwabe and Susanne Winkler (eds.), On Information Structure, Meaning and Form, –. Amsterdam: John Benjamins. Reichenbach, Hans (). Elements of Symbolic Logic. New York: Macmillan. Reilly, Jamie, Stacy Harnish, Amanda Garcia, Jinyi Hung, Amy D. Rodriguez, and Bruce Crosson (). Lesion symptom mapping of manipulable object naming in nonfluent aphasia: Can a brain be both embodied and disembodied? Cognitive Neuropsychology : –. Reilly, Judy, and Diane Anderson (). FACES: The acquisition of non-manual morphology in ASL. In Gary Morgan and Bencie Woll (eds.), Directions in Sign Language Acquisition, –. Amsterdam: Benjamins. Reinhart, Tanya (). Interface strategies. OTS Working Papers in Linguistics. Remberger, Eva-Maria (). Left peripheral interactions in Sardinian. Lingua : –. Repp, Sophie (). ¬(A&B): Gapping, negation and speech act operators. Research on Language and Computation  (): –. Repp, Sophie (). Common ground management: Modal particles, illocutionary negation and VERUM. In Daniel Gutzmann and Hans-Martin Gärtner (eds.), Beyond Expressives: Explorations in Use-Conditional Meaning, –. Leiden: Brill. Restall, Greg (). Multiple conclusions. In Petr Hájek, Luis Valdés-Villanueva, and Dag Westerståhl (eds.), Logic, Methodology, and Philosophy of Science: Proceedings of the Twelfth International Congress, –. London: Kings’ College Publications. Restall, Greg (). Truth values and proof theory. Studia Logica  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Restall, Greg (n.d.). Proof Theory and Meaning. Reuter, Tracy, Roman Feiman, and Jesse Snedeker (). Getting to no: Pragmatic and semantic factors in two- and three-year-olds’ understanding of negation. Child Development : e–e. Reverberi, Carlo, Luca L. Bonatti, Richard S. J. Frackowiak, Eraldo Paulesu, Paolo Cherubini, and Emiliano Macaluso (). Large scale brain activations predict reasoning profiles. Neuroimage : –. Reynolds, William T. (). Variation and Phonological Theory. Ph.D. dissertation, University of Pennsylvania. Rezac, Milan, Pablo Albizu, and Ricardo Etxepare (). The structural ergative of Basque and the theory of Case. Natural Language and Linguistic Theory : –. Richter, Frank, and Janina Radó (). Negative polarity in German: Some experimental results. Journal of Semantics : –. Richter, Frank, and Manfred Sailer (). LF constraints on expressions of Ty: An HPSG analysis of negative concord in Polish. In Robert D. Borsley and Adam Przepiórkowski (eds.), Slavic in HeadDriven Phrase Structure Grammar, –. Stanford, CA: CSLI Publications. Ripley, David (a). Contradictions at the borders. In R. Nouwen, R. van Rooij, U. Sauerland, and H.-C. Schmitz (eds.), Vagueness in Communication, –. Berlin: Springer. Ripley, David (b). Negation, denial, and rejection. Philosophy Compass  (): –. Ripley, David (). Paradoxes and failures of cut. Australasian Journal of Philosophy  (): –. Ripley, David (a). Anything goes. Topoi  (): –. Ripley, David (b). Embedding denial. In Colin Caret and Ole Hjortland (eds.), Foundations of Logical Consequence, –. Oxford: Oxford University Press. Ripley, David (). Bilateralism, coherence, warrant. In Friederike Moltmann and Mark Textor (eds.), Act-Based Conceptions of Propositional Content: Contemporary and Historical Perspectives, –. Oxford: Oxford University Press. Rispens, Judith, Roelien Bastiaanse, and Ron Van Zonneveld (). Negation in agrammatism: A cross-linguistic comparison. Journal of Neurolinguistics  (): –. Rizzi, Luigi (). Issues in Italian Syntax. Dordrecht: FORIS. Rizzi, Luigi (). Relativized Minimality. Cambridge, MA: MIT Press. Rizzi, Luigi (). The fine structure of the left periphery. In Liliane Haegeman (ed.), Elements of Grammar: Handbook in Generative Syntax, –. Dordrecht: Kluwer. Rizzi, Luigi (). Locality and left periphery. In Adriana Belletti (ed.), Structures and beyond: The Cartography of Syntactic Structures, –. Oxford: Oxford University Press. Roberts, Craige (). Context in dynamic interpretation. In Laurence R. Horn and Gregory Ward (eds.), The Handbook of Pragmatics, –. Oxford: Blackwell. Roberts, Craige (). Information structure in discourse: Towards an integrated formal theory of pragmatics. Semantics and Pragmatics : –. Roberts, Ian (). The C-system in Brythonic Celtic languages, V, and the EPP. In Luigi Rizzi (ed.), The Structure of CP and IP: The Cartography of Syntactic Structures, Vol. , –. Oxford: Oxford University Press. Roberts, Ian (). Principles and Parameters in a VSO Language: A Case Study in Welsh. Oxford: Oxford University Press. Roberts, Ian (). Diachronic Syntax. Oxford: Oxford University Press. Roberts, Ian (). Agreement and Head Movement: Clitics, Incorporation, and Defective Goals. Cambridge MA: MIT Press. Roberts, Ian, and Anna Roussou (). Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press. Roberts, William A., Miranda C. Feeney, Neil McMillan, Krista MacPherson, Evanya Musolino, and Mark Petter (). Do pigeons (Columba livia) study for a test? Journal of Experimental Psychology: Animal Behavior Processes : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Roelofsen, Floris, and Donka F. Farkas (). Polarity particle responses as a window onto the interpretation of questions and assertions. Language  (): –. Roelofsen, Floris, Noortja Verhuizen, and Galit Weidman Sassoon (). Positive and negative questions in discourse. In E. Chemla, V. Homer, G. Winterstein (eds.), Proceedings of Sinn und Bedeutung , –. Roeper, Peter (). Semantics for mass terms with quantifiers. Noûs  (): –. Roguska, Magdalena (). Exklamation und Negation. Ph.D. dissertation, Johannes Gutenberg Universität Mainz. Rohlfs, G. (–). Grammatica storica della lingua italiana e dei suoi dialetti. Turin: Einaudi. Romero, Maribel (). Biased yes/no questions: The role of . Sprache und Datenverarbeitung : –. Romero, Maribel (). High negation in subjunctive conditionals and polar questions. In Eva Csipak and Hedde Zeijlstra (eds.), Proceedings of Sinn und Bedeutung , –. Semanticsarchive.net. Romero, Maribel, and Chung-hye Han (). Verum focus in negative yes/no questions and Ladd’s p/¬p ambiguity. In Brendan Jackson (ed.), Proceedings of Semantics and Linguistic Theory XII, –. Ithaca, NY: CLC Publications. Romero, Maribel, and Chung-hye Han (). On negative yes/no questions. Linguistics and Philosophy : –. Romero, Maribel, Anja Arnhold, Bettina Braun, and Fillipo Domaneschi (). Negative polar question types in English. In Proceedings of the North East Linguistic Society , –. Amherst, MA: GLSA. Romoli, Jacopo (). The presuppositions of soft triggers are not presuppositions. In N. Ashton, A. Chereches, and D. Lut (eds.), Proceedings of Semantics and Linguistic Theory (SALT) . Romoli, Jacopo (). A scalar implicature-based approach to Neg-raising. Linguistics and Philosophy : –. Romoli, Jacopo (). The presuppositions of soft triggers are obligatory scalar implicatures. Journal of Semantics  (): –. Rooryck, Johan (). Negative and factive islands revisited. Journal of Linguistics : –. Rooth, Mats (). Association with Focus. Ph.D. dissertation, University of Massachusetts, Amherst. Rooth, Mats (). A theory of focus interpretation. Natural Language Semantics  (): –. Rooth, Mats (). Focus. In S. Lappin (ed.), The Handbook of Contemporary Semantic Theory, –. London: Blackwell. Rosemeyer, Malte (). The pragmatics of Spanish postposed-wh-interrogatives. Folia Linguistica. Rosemeyer, Malte, and Scott Schwenter (). Echoic and non-echoic affirmative responses in spoken Brazilian Portuguese. Ms. Ross, John R. (). Constraints on variables in syntax. Ph.D. dissertation, Massachusetts Institute of Technology. Ross, John R. (). Slifting. In M. Gross et al. (eds.), The Formal Analysis of Natural Languages, –. The Hague: Mouton. Ross, John R. (). Inner islands. In Proceedings of BLS . Berkeley: Berkeley Linguistic Society. Rothman, Jason (). Heritage speaker competence differences, language change, and input type: Inflected infinitives in Heritage Brazilian Portuguese. International Journal of Bilingualism  (): –. Roussou, Anna (). On the left periphery: Modal particles and complementisers. Journal of Greek Linguistics : –. Routley, Richard, and Valerie Routley (). The role of inconsistent and incomplete theories in the semantics of belief. Communication and Cognition : –. Rowlett, Paul (). Sentential Negation in French. Oxford: Oxford University Press. Rowlett, Paul (). Do French speakers really have two grammars? In B. Massot and P. Rowlett (eds.), L’hypothèse d’une diglossie en France. [Special issue, Journal of French Language Studies  ().]

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Rullmann, Hotze (a). Geen eenheid. Tabu : –. Rullmann, Hotze (b). Maximality in the semantics of WH-constructions. Ph.D. dissertation, University of Massachusetts, Amherst. Rullmann, Hotze (). Two types of negative polarity items. In Kiyomi Kusumoto (ed.), NELS : Proceedings of the North East Linguistic Society, –. Amherst: University of Massachusetts. Rullmann, Hotze (). Even, polarity, and scope. In M. Gibson, G. Wiebe, and G. Libben (eds.), Papers in Experimental and Theoretical Linguistics, –. University of Alberta, Edmonton, Canada. Rullmann, Hotze (). Additive particles and polarity. Journal of Semantics  (): –. Rumelhart, David (). Schemata: The building blocks of cognition. In R. Spiro, B. Bruce, and W. Brewer (eds.), Theoretical Issues in Reading Comprehension, –. Mahwah, NJ: Erlbaum Associates. Rumfitt, Ian (). ‘Yes’ and ‘No’. Mind  (): –. Rumfitt, Ian (). Asserting and excluding: Steps towards an anti-realist account of classical consequence. In Randall E. Auxier and Lewis Edwin Hahn (eds.), The Philosophy of Michael Dummett, –. Chicago: Open Court Press. Russ, Michael O., Wolfgang Mack, Carina-Raluca Grama, Heinrich Lanfermann, and Monika Knopf (). Enactment effect in memory: Evidence concerning the function of the supramarginal gyrus. Experimental Brain Research  (): –. Russell, Bertrand (). On denoting. Mind  (): –. Russell, Bertrand (). Inquiry into Meaning and Truth. New York: Allen & Unwin. Russell, Bertrand (). Human Knowledge: Its Scope and Limits. New York: Simon & Schuster. Ruys, E. G., and Y. Winter (). Quantifier scope in formal linguistics. In D. Gabbay and F. Guenther (eds.), Handbook of Philosophical Logic : –. Dordrecht: Springer. Sabbatini, Gloria, and Elisabetta Visalberghi (). Inferences about the location of food in capuchin monkeys (Cebus apella) in two sensory modalities. Journal of Comparative Psychology  (): –. Saddy, Douglas, Heiner Drenhaus, and Stefan Frisch (). Processing polarity items: Contrastive licensing costs. Brain and Language  (–); –. DOI: ./S-X ()-X. Sadock, Jerrold M. (). Queclaratives. In Douglas Adams, Mary Ann Campbell, Victor Cohen, Julie Lovins, Edward Maxwell, Carolyn Nygren, and John Reighard (eds.), Papers from the th Regional Meeting of the Chicago Linguistic Society, –. Chicago: CLS. Sadock, Jerrold M. (). Toward a Linguistic Theory of Speech Acts. New York: Academic Press. Sadock, Jerrold M. (). On testing for conversational implicature. In Peter Cole (ed.), Syntax and Semantics : Pragmatics, –. New York: Academic Press. Sadock, Jerrold M. (). Autolexical Syntax: A Theory of Parallel Grammatical Representations. Chicago: University of Chicago Press. Sag, Ivan (). Deletion and Logical Form. Massachusetts Institute of Technology, Ph.D. dissertation. Sailer, Manfred (). On reading-dependent licensing of strong negative polarity items. In Arndt Reister and Torgrim Solstad (eds.), Proceedings of Sinn und Bedeutung (SuB) , –. Stuttgart: University of Stuttgart. Sailor, Craig (). Questionable negation. Talk at the th Annual Meeting of the Linguistic Society of America. Boston, January , . Sainsbury, R. M. (). Facts and free logic. In Gerhard Preyer and Georg Peter (eds.), Protosociology: An International Journal of Interdisciplinary Research, Vol. : On Stephen Neale’s Facing Facts, –. Saltarelli, Mario (with Miren Azkarate, David Farwell, Jon Ortiz de Urbina, and Lourdes Oñederre) (). Basque. London: Croom Helm. Sambin, Giovanni, Guilia Battilotti, and Claudia Faggian (). Basic logic: Reflection, symmetry, visibility. Journal of Symbolic Logic  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Samek-Lodovici, Vieri (). The Interaction of Focus, Givenness, and Prosody. Oxford: Oxford University Press. Sandonato, Marie (). Zazaki. In Kahrel and ven den Berg (eds.), Typological Studies in Negation, –. Sanford, A. J., and A. C. Graesser (). Shallow processing and underspecification. Discourse Processes  (). –. Sanford, A. J., and P. Sturt (). Depth of processing in language comprehension: Not noticing the evidence. Trends in Cognitive Sciences  (): –. Sanford, A. J., L. M. Moxey, and K. B. Paterson (). Attentional focusing with quantifiers in production and comprehension. Memory and Cognition  (): –. Santi, Andrea and Yosef Grodzinsky (). Broca’s area and sentence comprehension: A relationship parasitic on dependency, displacement or predictability? Neuropsychologia : –. Santorini, Beatrice (). Variation and change in Yiddish subordinate clause word order. Natural Language and Linguistic Theory : –. Sapir, Edward (). Language: An Introduction to the Study of Speech. New York: Harcourt, Brace & Co. Sauerland, Uli (). The interpretation of traces. Natural Language Semantics : –. Sauerland, Uli, and Paul Elbourne (). Total reconstruction, PF movement, and derivational order. Linguistic Inquiry  (): –. Scappini, M., D. Delfitto, C. A. Marzi, F. Vespignani, and S. Savazzi (). On the non-incremental processing of negation: A pragmatically licensed sentence-picture verification study with Italian (dyslexic) adults. Cahiers de linguistique française : –. Scha, Remko (). Distributive, collective and cumulative quantification. In J. Groenendijk, Martin Stokhof, and Theo Janssen (eds.), Truth, Interpretation and Information. Dordrecht: Foris Publications. Schaeffer, Jeanette C. (). The Acquisition of Direct Object Scrambling and Clitic Placement: Syntax and Pragmatics. Amsterdam: John Benjamins. Schank, R. C., and R. P. Abelson (). Scripts, Plans, Goals and Understanding. Mahwah, NJ: Erlbaum Associates. Retrieved from . Scheffler, Tatjana (). Meaning variations in German tag-questions. Talk at the DGfS  Workshop “The prosody and meaning of (non-) canonical questions across languages”. Leipzig, March , . Schein, Barry (). Plurals and Events. Current Studies in Linguistics No. . Cambridge, MA: MIT Press. Schein, Barry (). Plurals. In Ernie Lepore and Barry C. Smith (eds.), The Oxford Handbook of Philosophy of Language, –. Oxford: Oxford University Press. Schein, Barry (). Event semantics. In Gillian Russell and Delia Graff Fara (eds.), The Routledge Companion to Philosophy of Language, –. London: Routledge. Schein, Barry (). Noughty bits: The subatomic scope of negation. Linguistics & Philosophy : –. Schein, Barry (). ‘And’: Conjunction Reduction Redux. Cambridge, MA.: The MIT Press. Schein, Barry (). Nip and tuck for definite description. Linguistics & Philosophy : –. Schiller, N. O., L. van Lenteren, J. Witteman, K. Ouwehand, G. P. Band, and A. Verhagen (). Solving the problem of double negation is not impossible: Electrophysiological evidence for the cohesive function of sentential negation. Language, Cognition and Neuroscience : –. Schimke, Sarah, and Christine Dimroth (). The influence of finiteness and lightness on verb placement in L German: Comparing child and adult learners. Second Language Research  (): –. Schindele, R., J. Lüdtke, and B. Kaup (). Comprehending negation: A study with adults diagnosed with high functioning autism or Asperger’s syndrome. Journal of Intercultural Pragmatics : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Schlenker, Philippe (). Be articulate: A pragmatic theory of presupposition projection. Theoretical Linguistics  (): –. Schlenker, Philippe (). Local contexts. Semantics and Pragmatics : –. Schlenker, Philippe (). Super monsters II: Role Shift, iconicity and quotation in sign language. Semantics and Pragmatics  (): –. Schloegl, Christian, Thomas Bugnyar, and Ulrike Aust (). Exclusion performances in nonhuman animals: From chimpanzees to pigeons and back again. In A. Blaisdell, L. Huber, S. Watanabe, A. Young, and Y. Yamazaki (eds.), Rational Animals, Irrational Humans, –. Tokyo: Keio University Press. Schloegl, Christian, Judith Schmidt, Markus Boeckle, Brigitte M. Weiß, and Kurt Kotrschal (). Grey parrots use inferential reasoning based on acoustic cues alone. Proc. R. Soc. B : –. Schmelz, Martin, Josep Call, and Michael Tomasello (). Chimpanzees know that others make inferences. Proceedings of the National Academy of Sciences : –. Schmerling, Susan (). A note on negative polarity. Papers in Linguistics : –. Schroeder-Heister, Peter (). The categorical and the hypothetical: A critique of some fundamental assumptions of standard semantics. Synthese  (): –. Schütze, Carson (). INFL in child and adult language: Agreement, case and licensing. Ph.D. dissertation, MIT. Schütze, Carson (). The status of he/she don’t and theories of root infinitives. (accessed September ). Schütze, Carson (). The status of non-agreeing don’t and theories of root infinitives. Language Acquisition : –. Schwartz, Bonnie (). Child L acquisition: Paving the way. In: B. Beachley, A. Brown, and F. Conlin (eds.), Proceedings of the th Annual BUCLD, Vol. , –. Somerville, MA: Cascadilla Press. Schwartz, Bonnie (). Why child L acquisition? In J. Van Kampen, and S. Baauw (eds.), Proceedings of Generative Approaches to Language Acquisition Conference . LOT Occasional Series , –. Schwartz, Bonnie, and Rex Sprouse (). Word order and nominative case in non-native language acquisition: A longitudinal study of (L Turkish) German Interlanguage. In T. Hoekstra and B. Schwartz (eds.), Language Acquisition Studies in Generative Grammar, –. Amsterdam: John Benjamins. Schwartz, Bonnie, and Rex Sprouse (). L cognitive states and the Full Transfer/Full Access model. Second Language Research  (): –. Schwarz, Bernhard (). Scalar additive particles in negative contexts. Natural Language Semantics : –. Schwarzschild, Roger (). Plurals, presuppositions and the sources of distributivity. Natural Language Semantics  (): –. Schwarzschild, Roger (). Givenness, avoid F and other constraints on the placement of accent. Natural Language Semantics : –. Schwarzschild, Roger (). The semantics of comparatives and other degree constructions. Language and Linguistics Compass : –. Schwarzschild, Roger, and Karina Wilkinson (). Quantifiers in comparatives: A semantics of degree based on intervals. Natural Language Semantics : –. Schwegler, Armin (). Predicate negation and word-order change: A problem of multiple causation. Lingua : –. Schwegler, Armin (). Word-order changes in predicate negation strategies in Romance languages. Diachronica : –. Schwenter, Scott A. (). The pragmatics of negation in Brazilian Portuguese. Lingua : –. Schwenter, Scott A. (). Fine tuning Jespersen’s cycle. In Betty J. Birner and Gregory. L. Ward (eds.), Drawing the Boundaries of Meaning: Neo-Gricean Studies in Pragmatics and Semantics in Honor of Laurence R. Horn, –. Amsterdam: John Benjamins.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Schwenter, Scott A., and Mary Johnson (). Neg Nada Constructions in Comparative Perspective: Brazilian Portuguese and Argentinian Spanish. Presentation at th IPRA, July –, , University of Manchester. Powerpoint presentation,  slides. Schwenter, Scott A., Kendra Dickinson, and Luana Lamberti (). Pragmatic strength and (lack of) negative concord. Poster, rd annual meeting, Linguistic Society of America. Searle, John R. (). Speech Acts: An Essay in the Philosophy of Language. Cambridge: Cambridge University Press. Searle, John R. (). Expression and Meaning: Studies in the Theory of Speech Act. Cambridge: Cambridge University Press. Sedivy, Julie (). Against a unified analysis of negative polarity licensing. Cahiers de Linguistique de Ottawa : –. Sells, Peter (). Interactions of negative polarity items in Korean. Proceedings of the th Harvard International Symposium on Korean Linguistics, –. Cambridge, MA: Harvard University. Sells, Peter, and Shin-Sook Kim (). Korean NPIs scope over negation. Language Research : –. Selvanayagam, J., V. Witte, L. A. Schmidt, and V. D. Dwivedi (). A preliminary investigation of dispositional affect, the P, and sentence processing. Brain Research : –. Servidio, Emilo (). Contrastive topics accommodate antecedents for ellipsis. Talk presented at the Identity in Ellipsis workshop, Universiteit Leiden, September –. Servidio, Emilio, Giuliano Bocci, and Valentina Bianchi (). (Dis)agreement, polarity, and focus: Answering negative polar questions in Italian. Glossa: A Journal of General Linguistics  (): .–. Seuren, Pieter (). Negation in Dutch. Neophilologus : –. Seuren, Pieter (). The comparative. In Ference Kiefer and Nicolas Ruwet (eds.), Generative Grammar in Europe, –. Dordrecht: Reidel. Seuren, Pieter (). Negative’s travels. In P. A. M. Seuren (ed.), Semantic Syntax. Oxford: Oxford University Press. Seuren, Pieter (). Meer over minder dan hoeft. De nieuwe taal. Seuren, Pieter (). Burton-Roberts on presupposition and negation. Journal of Linguistics : –. Seuren, Pieter (). The natural logic of language and cognition. Pragmatics : –. Seuren, Pieter, and Dany Jaspers (). Logico-cognitive structure in the lexicon. Language : –. Severens, E., B. M. Jansma, and R. J. Hartsuiker (). Morphophonological influences on the comprehension of subject–verb agreement: An ERP study. Brain Research : –. Shapiro, Stewart (). Simple truth, contradiction, and consistency. In Graham Priest, J. C. Beall, and Bradley Armour-Garb (eds.), The Law of Non-Contradiction. Oxford: Oxford University Press. Sheehan, Michelle (). The Final-over-Final Condition and adverbs. In Michelle Sheehan, Theresa Biberauer, Anders Holmberg, and Ian Roberts (eds.), The Final-over-Final Condition: A Syntactic Universal, –. Cambridge, MA: MIT Press. Sherman, Mark A. (). Adjectival negation and the comprehension of multiply negated sentences. Journal of Verbal Learning and Verbal Behavior  (): –. Shimoyama, Junko (). Indeterminate phrase quantification in Japanese. Natural Language Semantics : –. Shimoyama. Junko (). Japanese indeterminate negative polarity items and their scope. Journal of Semantics : –. Shlonsky, Ur (). Clause Structure and Word Order in Hebrew and Arabic. An Essay in Comparative Semitic Syntax. Oxford: Oxford University Press. Shtyrov, Yury, Anna Butorina, Anastasia Nikolaeva, and Tatiana Stroganova (). Automatic ultrarapid activation and inhibition of cortical motor systems in spoken word comprehension. Proceedings of the National Academy of Sciences : E–E. Shyu, Shu-ing (). Minimizers and EVEN. Linguistics  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Simon, H. A. (). Rational choice and the structure of the environment. Psychological Review  (): . Simons, Mandy (). On the conversational basis of some presuppositions. In R. Hastings, B. Jackson, and S. Zvolenszky (eds.), Semantics and Linguistic Theory, . Ithaca, NY: CLC Publications. Simons, Mandy (). Presupposition without common ground. Unpublished manuscript, Carnegie Mellon University. Simons, M., J. Tonhauser, D. Beaver, and C. Roberts (). What projects and why. Semantics and Linguistic Theory : –. Simpson, Andrew, and Zoe Wu (). Agreement, shells and focus. Language  (): –. Sitaridou, Ioanna (). Modality, antiveridicality and complementation: The Romeyka infinitive as a negative polarity item. Lingua : –. Sjörs, Ambjörn (). Historical Aspects of Standard Negation in Semitic. Leiden: Brill. Skopeteas, Stavros, and Gisbert Fanselow (). Focus and the exclusion of alternatives: On the interaction of syntactic structure with pragmatic inference. Lingua  (): –. Slabakova, Roumyana (). Meaning in the Second Language. Berlin: Walter de Gruyter. Sloggett, Shayne (). When Errors Aren’t: How Comprehenders Selectively Violate Binding Theory (Ph.D. dissertation, University of Massachusetts Amherst). Smessaert, Hans, and Lorenz Demey (). Logical geometries and information in the Square of Oppositions. Journal of Logic, Language and Information : –. Smiley, Timothy (). Rejection. Analysis  (): –. Smith, Jennifer (). Negative concord in the Old and New World: Evidence from Scotland. Language Variation and Change : –. Smith, J. L., S. J. Johnstone, and R. J. Barry (). Movement-related potentials in the Go/NoGo task The P reflects both cognitive and motor inhibition. Clin Neurophysiol : –. Smith, L., and C. Yu (). Infants rapidly learn word-referent mappings via cross-situational statistics. Cognition  (): –. Smith, Susan M. (). Black-Capped Chickadee. Mechanicsburg, PA: Stackpole Books. Soames, Scott (). A projection problem for speaker presuppositions. Linguistic Inquiry  (): –. Soames, Scott (). How presuppositions are inherited: A solution to the projection problem. Linguistic Inquiry  (): –. Solomon, R. L., and L. C. Wynne (). Traumatic avoidance learning: Acquisition in normal dogs. Psychological Monographs: General and Applied  (): . Sorace, Antonella (). Near-nativeness. In C. Doughty and M. Long (eds.), Handbook of Second Language Acquisition, –. Oxford: Blackwell. Sorace, Antonella (). Selective optionality in language development. In L. Cornips and K. Corrigan (eds.), Syntax and Variation: Reconciling the Biological and the Social, –. Amsterdam: John Benjamins. Sorace, Antonella (). Pinning down the concept of “interface” in bilingualism. Linguistic Approaches to Bilingualism  (): –. Sorace, Antonella, and Francesca Filiaci (). Anaphora resolution in near-native speakers of Italian. Second Language Research  (): –. Sorace, Antonella, and Ludovica Serratrice (). Internal and external interfaces in bilingual language development: Beyond structural overlap. International Journal of Bilingualism  (): –. Sorensen, Roy (). Vagueness. In E. Zalta (ed.), Stanford Encyclopedia of Philosophy. . Speas, Peggy, and Carol Tenny (). Configurational properties of point of view roles. In Anna Maria Di Sciullo (ed.), Asymmetry in Grammar, Vol. : Syntax and Semantics, –. Amsterdam: John Benjamins. Spector, Benjamin (). Homogeneity and plurals: From the Strongest Meaning Hypothesis to Supervaluations. Sinn und Bedeutung .

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Spenader, Jennifer, and Emar Maier (). Contrast as denial in multi-dimensional semantics. Journal of Pragmatics  (): –. Sperber, D., and D. Wilson (). Relevance: Communication and Cognition. Oxford: Blackwell. Sperber, D., and D. Wilson (). Relevance: Communication and Cognition. Oxford/Cambridge: Blackwell Publishers (nd edn.). Squartini, Mario (). Italian non-canonical negations as modal particles: Information state, polarity and mirativity. In Chiara Fedriani and Andrea Sansó (eds.), Pragmatic Markers, Discourse Markers and Modal Particles: New Perspectives, –. Amsterdam: Benjamins. Sridhar, S. N. (). Kannada. London: Routledge. Stainton, Robert J. (). Quantifier phrases, meaningfulness “in isolation,” and ellipsis. Linguistics and Philosophy : –. Stainton, Robert J. (). In defense of non-sentential assertion. In Zoltán Gendler Szabó (ed.), Semantics Versus Pragmatics, –. Oxford: Oxford University Press. Stainton, Robert J. (a). Neither fragments nor ellipsis. In Ljiljana Progovac, Kate Paesania, Eugenia Casielles, and Ellen Barton (eds.), The Syntax of Nonsententials, –. Amsterdam/ New York: John Benjamins. Stainton, Robert J. (b). Words and Thoughts: Subsentences, Ellipsis, and the Philosophy of Language. Oxford: Oxford University Press. Stalnaker, Robert (). Pragmatics. In D. Davidson and G. Harman (eds.), Semantics of Natural Language, –. Reidel: Dordrecht. Stalnaker, Robert (). Presuppositions. Journal of Philosophical Logic  (): –. Stalnaker, Robert (). Pragmatic presuppositions. In M. Munitz and P. Unger (eds.), Semantics and Philosophy, –. Columbia: New York University Press. Stalnaker, Robert (). Assertion. Syntax and Semantics : –. New York: Academic Press. Stalnaker, Robert (). Common ground. Linguistics and Philosophy : –. Stanley, Jason (). Context and logical form. Linguistics and Philosophy : –. Stark, Elisabeth (). Indefinitheit und Textkohärenz: Entstehung und Semantische Strukturierung Indefiniter Nominaldetermination im Altitalienischen. Tübingen: Niemeyer. Starke, Michal (). Nanosyntax: A short primer to a new approach to language. Nordlyd : –. Starke, Michal (). Towards an elegant solution to language variation: Variation reduces to the size of lexically stored trees. Ms., Tromsø University. Stauble, A. M. (). The process of decreolization: A model for second language development. Language Learning  (): –. Stauble, A. M. (). A comparison of a Spanish–English and a Japanese–English second language continuum: Negation and verb morphology. In R. Andersen (ed.), Second Languages: A CrossLinguistic Perspective. Rowley, MA: Newbury House. Steedman, Mark (). Taking Scope. Cambridge, MA: MIT Press. Steinhauer, K., E. J. White, and J. E. Drury (). Temporal dynamics of late second language acquisition: Evidence from event-related potentials. Second Language Research : –. Steinhauer, K., J. E. Drury, P. Portner, M. Walenski, and M. Ullmann (). Syntax, concepts and logic in the temporal dynamics of language comprehension: Evidence from Event Related Potentials. Neuropsychologia : –. St. George, M., S. Mannes, and J. Hoffman, (). Individual differences in inference generation: An ERP analysis. J. Cogn. Neurosci. : –. Stickel, Gerhard (). Untersuchungen zur Negation im heutigen Deutsch. Braunschweig: Vieweg. Stojanovich, Isadora (). Situation semantics. In A. Newen and R. van Riel (eds.), Identity, Language and Mind: An Introduction to the Philosophy of John Perry, –. Stanford, CA: CSLI Publications. Storto, Luciana (). Negation in Karitiana. In Lisa Matthewson, Erin Guntly, Marianne Huijsmans, and Michael Rochemont (eds.), Wa xweysás i nqwal’utteníha i ucwalmícwa: He Loves the People’s Language. Essays in Honour of Henry Davis, –. University of British Columbia Occasional Papers in Linguistics. Strawson, P. F. (). On referring. Mind  (): –. Strawson, P. F. (). Introduction to Logical Theory. London: Methuen & Co. Ltd.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Strawson, P. F. (). Identifying reference and truth values. Theoria  (): –. Stromswold, Karin (). Learnability and the Acquisition of Auxiliaries. Ph.D. dissertation, MIT. Sturt, Patrick (). The time-course of the application of binding constraints in reference resolution. Journal of Memory and Language  (): –. Sudo, Yasutada (). Biased polar questions in English and Japanese. In Daniel Gutzmann and Hans-Martin Gärtner (eds.), Beyond Expressives: Explorations in Use-Conditional Meaning, –. Leiden: Brill. Süleyman, Ulutaş (). Feature inheritance and subject Case in Turkish. In S. Ay et al. (eds.), Essays on Turkish Linguistics, –. Wiesbaden: Harrassowitz. Suleymanova, Venera, and Jack Hoeksema (). Minimizers in Azerbaijani from a comparative perspective. Folia Linguistica  (): –. Surányi, Balázs (). Quantification and focus in negative concord. Lingua  (): –. Svartvik, Jan (ed.) (). The London Corpus of Spoken English: Description and Research. Lund Studies in English, . Lund: Lund University Press. Svartvik, Jan, and Randolph Quirk (eds.) (). A Corpus of English Conversation, Lund Studies in English, . Lund: Liber/Gleerups. Swets, B., T. Desmet, C. Clifton, and F. Ferreira (). Underspecification of syntactic ambiguities: Evidence from self-paced reading. Memory and Cognition  (): –. Ph.D. Swingley, Daniel (). Fast mapping and slow mapping in children’s word learning. Language Learning and Development  (): –. Szabolcsi, Anna (). The semantics of topic/focus articulation. In Jeroen Groenendijk, Theo Janssen, and Martin Stokhof (eds.), Formal Methods in the Study of Language, –. Amsterdam: Mathematical Center. Szabolcsi, Anna (). Positive polarity—negative polarity. Natural Language and Linguistic Theory : –. Szabolcsi, Anna (). Quantification. Cambridge: Cambridge University Press. Szabolcsi, Anna, and Bill Haddican (). Conjunction meets negation: A study of cross-linguistic variation. Journal of Semantics  (): –. Szabolcsi, Anna, and Frans Zwarts (). Weak islands and an algebraic semantics for scope taking. Natural Language Semantics : –. Szymanik, Jakub (). Quantifiers and Cognition: Logical and Computational Perspectives. Studies in Linguistics and Philosophy. Berlin: Springer. Tabor, W., B. Galantucci, and D. Richardson (). Effects of merely local syntactic coherence on sentence processing. Journal of Memory and Language  (): –. Takahashi, Shoichi (). More than two quantifiers. Natural Language Semantics : –. Tam, Clara W-Y., and Stephanie F. Stokes (). Form and function of negation in early developmental Cantonese. Journal of Child Language : –. Tanenhaus, Michael K., Michael J. Spivey-Knowlton, Kathleen M. Eberhard, and Julie C. Sedivy (). Integration of visual and linguistic information in spoken language comprehension. Science : –. Tang, Gladys(). Questions and negation in Hong Kong Sign Language. In Ulrike Zeshan (ed.), Interrogative and Negative Constructions in Sign Languages, –. Nijmegen: Ishara Press. Tanner, D. (). Robust neurocognitive individual differences in grammatical agreement processing: A latent variable approach. Cortex : –. Tanner, D., and J. van Hell (). ERPs reveal individual differences in morphosyntactic processing. Neuropsychologia : –. Tanner, D., J. McLaughlin, J. Herschensohn, and L. Osterhout (). Individual differences reveal stages of L grammatical acquisition: ERP evidence. Bilingualism: Language and Cognition : –. Tappenden, Jamie (). Negation, denial, and language change in philosophical logic. In Dov M. Gabbay and Heinrich Wansing (eds.), What Is Negation?, –. Dordrecht: Kluwer Academic Publishers.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Tauli, Valter (). Standard Estonian Grammar: Syntax. Uppsala: Acta Universitatis Upsaliensis. Taylor, Ann, Anthony Warner, Susan Pintzuk, and Frank Beths (eds.) (). The York-TorontoHelsinki Parsed Corpus of Old English Prose. York: Department of Language and Linguistic Science, University of York. Taylor, Barry (). Modes of Occurrence. Oxford: Basil Blackwell. Temmerman, Tanja (). Multidominance, Ellipsis, and Quantifier Scope. Ph.D. dissertation. Leiden University. Temmerman, Tanja (). The syntax of Dutch embedded fragment answers: On the PF-theory of islands and the /sluicing correlation. Natural Language and Linguistic Theory  (): –. Tennant, Neil (). Negation, absurdity, and contrariety. In Dov M. Gabbay and Heinrich Wansing (eds.), What Is Negation?, –. Dordrecht: Kluwer Academic Publishers. Tesan, Graciela, Blake W. Johnson, and Stephen Crain (). How the brain responds to any: An MEG study. Brain and Language  (): –. DOI: ./j.bandl.... Tettamanti, Marco, Giovanni Buccino, Maria Cristina Saccuman, Vittorio Gallese, Massimo Danna, Paola Scifo, Ferruccio Fazio, Giacomo Rizzolatti, Stefano F. Cappa, and Daniela Perani (). Listening to action-related sentences activates fronto-parietal motor circuits. Journal of Cognitive Neuroscience : –. Tettamanti, Marco, Rosa Manenti, Pasquale A. Della Rosa, Andrea Falini, Daniela Perani, Stefano F. Cappa, and Andrea Moro (). Negation in the brain: Modulating action representations. NeuroImage  (): –. Textor, Mark (). Is ‘no’ a force-indicator? No! Analysis  (): –. Thompson, Laurence C. (). A Vietnamese Grammar. Seattle: University of Washington Press. Thoms, Gary (). Syntactic identity, Parallelism and accommodated antecedents. Lingua : –. Thornton, Rosalind, and Kelly Rombough (). The syntax–PF interface in children’s negative sentences. Language Acquisition : –. Thornton, Rosalind, and Graciela Tesan (). Sentential negation in early child English. Journal of Linguistics : –. Thornton, Rosalind, Anna Notley, Vincenzo Moscati, and Stephen Crain (). Two negations for the price of one. Glossa: A Journal of General Linguistics  (): .–. Thurgood, Graham (). From Ancient Cham to Modern Dialects: Two Thousand Years of Language Contact and Change. Honolulu: University of Hawai’i Press. Tian, Ye, and Richard Breheny (). Dynamic pragmatic view of negation processing. In P. Larrivée and C. Lee (eds.), Negation and Polarity: Experimental Perspectives, –. Berlin: Springer. Tian, Ye, Richard Breheny, and Heather J. Ferguson (). Why we simulate negated information: A dynamic pragmatic account. The Quarterly Journal of Experimental Psychology : –. Tiedemann, Jörg (). Parallel data, tools and interfaces in opus. In Nicoletta Calzolari et al. (eds.), Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC), –. Istanbul: European Language Resources Association (ELRA). Tieu, Lyn (). On the tri-ambiguous status of any: The view from child language. In D. Lutz and N. Li (eds.), Proceedings of the th Semantics and Linguistic Theory Conference, –. Ithaca, NY: CLC Publications, Cornell University. Tieu, Lyn (). Logic and Grammar in Child Language. Ph.D. dissertation, University of Connecticut. Tieu, Lyn (). Input versus output in the acquisition of negative polarity: The curious case of any. In Pierre Larrivée and Chungmin Lee (eds.), Negation and Polarity: Experimental Perspectives, Language Cognition, and Mind . Tieu, Lyn, and Jeffrey Lidz (). NPI licensing and beyond: Children’s knowledge of the semantics of any. Language Acquisition : –. Tobler, Adolf (). Il ne faut pas que tu meures “du darfst nicht sterben”. Reprinted in Vermischte Beiträge zur französischen Grammatik I, rd edn., –. Leipzig: S. Hirzel. Tomaselli, Alessandra, and Bonnie Schwartz (). Analysing the acquisition stages of negation in L German: Support for UG in adult SLA. Second Language Research  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Tomasello, Michael (). Origins of Human Communication. Cambridge, MA: MIT Press. Tomasino, Barbara, Peter H. Weiss, and Gereon R. Fink (). To move or not to move: Imperatives modulate action-related verb processing in the motor system. Neuroscience : –. Tomioka, Satoshi (). Pragmatics of LF intervention effects: Japanese and Korean Whinterrogatives. Journal of Pragmatics : – Tonhauser, J., D. Beaver, C. Roberts, and M. Simons (). Toward a taxonomy of projective content. Language  () –. Toosarvandani, Maziar (). Corrective but coordinates clauses not always but sometimes. Natural Language and Linguistic Theory  (): –. Tortora, Christina, and Marcel den Dikken (). Subject agreement variation: Support for the configurational approach. Lingua : –. Tortora, Christina, Beatrice Santorini, Frances Blanchette, and C. E. A. Diertani (). The AudioAligned and Parsed Corpus of Appalachian English (AAPCAppE), version .. . Tottie, Gunnel (a). Lexical diffusion in syntactic change: Frequency as a determinant of linguistic conservatism in the development of negation in English. In Dieter Kastovsky (ed.), Historical English Syntax, –. Berlin: Mouton de Gruyter. Tottie, Gunnel (b). Negation in English Speech and Writing: A Study in Variation. Quantitative analyses of linguistic structure. San Diego: Academic Press. Tottie, Gunnel, and Anja Neukom-Hermann (). Quantifier–negation interaction in English: A corpus linguistic study of all . . . not constructions. In Horn (ed.) The Expression of Negation, –. Tovena, Lucia M. (). The fine structure of polarity sensitivity. Outstanding dissertations in linguistics. New York: Garland Publishing Inc. Tovena, Lucia M. (a). Neg-raising: Negation as failure? In J. Hoeksema, H. Rullmann, V. Sánchez Valencia, and T. van der Wouden (eds.), Perspectives on Negation and Polarity Items, –. Amsterdam: John Benjamins. Tovena, Lucia M. (b). The phenomenon of polarity sensitivity: Questions and answers. Lingua e Stile XXXVI, –. Tovena, Lucia M., and J. Jayez (). Any: From scalarity to arbitrariness. In F. Corblin, J.M. Marandin, and C. D.-Sorin (eds.), Empirical Issues in Formal Syntax and Semantics , –. The Hague: Holland Academic Graphic. Tovena, Lucia M., Viviane Déprez, and Jacques Jayez (). Polarity sensitive items. In F. Corblin and H. de Swart (eds.), Handbook of French Semantics, –. Stanford, CA: CSLI Publications. Townsend, D. J., and T. G. Bever (). Sentence Comprehension: The Integration of Rules and Habits. Cambridge, MA: MIT Press. Trabasso, T., H. Rollins, and E. Shaughnessy (). Storage and verification stages in processing concepts. Cognitive Psychology : –. Trueswell, J. C., T. N. Medina, A. Hafri, and L. R. Gleitman (). Propose but verify: Fast mapping meets cross-situational word learning. Cognitive Psychology  (): –. Tsimpli, Ianthi-Maria, and Anna Roussou (). Negation and polarity items in Greek. The Linguistic Review : –. Tubau, Susagna (). Syntax–Morphology Interface Conditions on the Expression of Negation. Ph.D. dissertation, Universitat Autònoma Barcelona and University of Amsterdam. Tubau, Susagna (a). Lexical variation and Negative Concord in traditional dialects of British English. The Journal of Comparative Germanic Linguistics : –. Tubau, Susagna (b). On the syntax of English minimizers. Natural Language and Linguistic Theory  (): –. Tubau, Susagna, Viviane Déprez, J. Borràs-Comes, and M. T. Espinal (). How speakers interpret the negative markers no and no . . . pas in Catalan. Probus  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Tubau, Susagna, Santiago González-Fuente, Pilar Prieto, and M.Teresa Espinal (). Prosody and gesture in the interpretation of yes-answers to negative yes-no questions. In J. Fortuny and B. Coromines (eds.), On the Locus of Ambiguity and the Design of Language. Thematic issue of The Linguistic Review  (): –. Tunstall, Susanne (). The interpretation of quantifiers: Semantics and processing. Ph.D. dissertation, University of Massachusetts-Amherst. Turner, Ken (). Defending semantic presupposition. Journal of Pragmatics  (): –. Ulrich, R., and C. Maienborn (). Left-right coding of past and future in language: The mental timeline during sentence processing. Cognition : –. Unsworth, Sharon (). Young (and older) L learners and Dutch scrambling. In M. Cazzoli Goeta, S. Pourcel, and L. Van Espen (eds.), Proceedings of the Fifth Durham Postgraduate Conference in Theoretical and Applied Linguistics, –. Durham: School of Linguistics and Language, University of Durham. Unsworth, Sharon (). Child L, child L and adult L acquisition: Differences and similarities. A study on the acquisition of direct object scrambling in Dutch. Unpublished doctoral dissertation, Utrecht University, the Netherlands. Unsworth, Sharon (). Comparing child L development with adult L development: How to measure L proficiency. In E. Gavruseva and B. Haznedar (eds.), Current Trends in Child Second Language Acquisition: A Generative Perspective, –. Amsterdam/Philadelphia: John Benjamins. Ura, Hiroyuki (). Multiple feature-checking: A theory of grammatical function splitting. Ph.D. dissertation, MIT. Uribe-Echevarria, Maria (). Interface licensing conditions on negative polarity items: A theory of polarity and tense interactions. Ph.D. dissertation, University of Connecticut. Vai, Massimo (). Per una storia della negazione in Milanese in comparazione con altre varietà altoitaliane. ACME : –. Vainikka, Anne, and Martha Young-Scholten (). Gradual development of L phrase structure Second Language Research  (): –. Vallduví, Enric (). Polarity items, n-words and minimizers in Catalan and Spanish. Probus : –. Valmala, Vidal (). The syntax of little things. Paper presented at the th Colloquium on Generative Grammar, Girona, June . . Last accessed January , . Van Alsenoy, Lauren (). A New Typology of Indefinite Pronouns, with a Focus on Negative Indefinites. Ph.D. dissertation, University of Antwerp. Van Baar, Tim (). Phasal Polarity. Ph.D. dissertation, University of Amsterdam. van Berkum, J. J. A., P. Hagoort, and C. M. Brown (). Semantic integration in sentences and discourse: Evidence from the N. J. Cogn. Neurosci. : –. van Berkum, J. J. A., P. Zwitserlood, P. Hagoort, and C. M. Brown (). When and how do listeners relate a sentence to the wider discourse? Evidence from the N effect. Cogn. Brain Res. : –. van Craenenbroeck, Jeroen (). Ellipsis, identity, and accommodation. Ms., Katholieke Universiteit Leuven/Hogeschool-Universiteit Brussel. . Last accessed January , . van Craenenbroeck, Jeroen, and Jason Merchant (). Ellipsis phenomena. In Marcel den Dikken (ed.), The Cambridge Handbook of Generative Syntax, –. Cambridge: Cambridge University Press. Vandamme, Fernand (). On negation, an interdisciplinary study. Logique et analyse : –. Van den Heuvel, Wilco (). Biak, Description of an Austronesian Language of Papua. Ph.D. dissertation, Free University of Amsterdam. van der Auwera, Johan (). ‘Already’ and ‘still’: Beyond duality. Linguistics and Philosophy : –. van der Auwera, Johan (). Modality: The three-layered scalar square. Journal of Semantics : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





van der Auwera, Johan (). On the typology of negative modals. In J. Hoeksema, H. Rullmann, V. Sanchez-Valencia, and T. van der Wouden (eds.), Perspectives on Negation and Polarity Items, –. Amsterdam: John Benjamins. van der Auwera, Johan (). Why languages prefer prohibitives. 外国语 (Wai guo yu—Journal of Foreign Languages) : –. van der Auwera, Johan (). The Jespersen cycles. In E. van Gelderen (ed.), Cyclical Change, –. Amsterdam: Benjamins. van der Auwera, Johan (a). Prohibition: constructions and markers. In D. Shu et al. (eds.), Contrasting Meaning in Languages of the East and the West, –. Bern: Peter Lang. van der Auwera, Johan (b). The diachrony of negation. In Laurence R. Horn (ed.), The Expression of Negation, –. Berlin/New York: de Gruyter Mouton. van der Auwera, Johan (). Standard Average European. In Bernd Kortmann and Johan van der Auwera (eds.), The Languages and Linguistics of Europe: A Comprehensive Guide, –. Berlin: de Gruyter Mouton. van der Auwera, Johan (). Wat het Nederlands ons kan leren over de Jespersencyclus [What Dutch can teach us about the Jespersen Cycle]. Nederlandse Taalkunde : –. van der Auwera, Johan (). More ado about *nall. In J. Blochowiak, C. Grisot, S. DurrlemannTame, and C. Laenzlinger (eds.), Collection of Papers Dedicated to Jacques Moeschler. . van der Auwera, Johan (). Englishes, English Creoles and their negative indefinites. In Debra Ziegeler and Zhiming Bao (eds.), Negation and Contact: With Special Reference to Singapore English, –. Amsterdam: Benjamins. van der Auwera, Johan, and D. N. S. Bhatt (). Modality and negation in Havyaka Kannada. Indian Linguistics : –. van der Auwera, Johan, and Paul Gybels (). On negation, indefinites and negative indefinites in Yiddish. In Marion Aptroot (ed.), Yiddish Language Structures, –. Berlin: de Gruyter. van der Auwera, Johan, and Tamirand Nnena de Lisser (). Negative concord in Jamaican Ampersand . . van der Auwera, Johan, and Lauren van Alsenoy (). On the typology of Negative Concord. Studies in Language  (): –. van der Auwera, Johan, and Lauren van Alsenoy (). More ado about nothing: On the typology of negative indefinites. In Ken Turner and Laurence R. Horn (eds.), Pragmatics, Truth and Underspecification: Towards an Atlas of Meaning, –. Leiden: Brill. van der Auwera, Johan, and Frens Vossen (). Negatives between Chamic and Bahnaric’, Journal of the Southeast Asian Linguistics Society : –. van der Auwera, Johan, and Frens Vossen (). Jespersen Cycles in the Mayan, Quechuan and Maipurean languages. In E. Van Gelderen (ed.), Cyclical Change Continued, –. Amsterdam: Benjamins. van der Auwera, Johan, and Frens Vossen (). Kiranti double negation: A copula conjecture. Lingistics of the Tibeto-Burman Area : –. van der Auwera, Johan, Ludovic Decuypere, and Annemie Neuckermans (). Negative indefinites: A typological and diachronic perspective on a Brabantic construction. In Terttu Nevalainen, Juhani Klemola, and Miko Laitinen (eds.), Types of Variation: Diachronic, Dialectal and Typological Interfaces, –. Amsterdam: Benjamins. van der Auwera, Johan, Olga Krasnoukhova, and Frens Vossen (forthcoming). Intertwining the negative cycles. In Ljuba Veselinova and Arja Hamari (eds.), The Negative Existential Cycle from a Historical-Comparative Perspective. Berlin: Language Science Press. van der Auwera, Johan, and Ludo Lejeune with Valentin Goussev (). The prohibitive. In Dryer and Haspelmath (eds.), , last accessed on August , . van der Auwera, Johan, Frens Vossen, and Maud Devos (). Le cycle de Jespersen à trois et quatre négations. In F. Jacques (ed.), La linguistique de la contradiction, –. Brussels: Lang.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





van der Does, Jaap (). A generalized quantifier logic for naked infinitives. Linguistics & Philosophy : –. van der Sandt, Rob A. (). Kontekst en presuppositie: een studie van het projektieprobleem en de presuppositionele eigenschappen van de logische konnektieven. Ph.D. dissertation, Nijmegen: Nijmegen Institute of Semantics. van der Sandt, Rob A. (). Context and Presupposition. London: Croom Helm. van der Sandt, Rob A. (). Denial. Chicago Linguistic Society : The Parasession on Negation, –. van der Sandt, Rob A. (). Presupposition projection as anaphora resolution. Journal of Semantics  (): –. van der Sandt, Rob A. (). Denial and presupposition. In Peter Kühnlein, Hannes Rieser, and Henk Zeevat (eds.), Perspectives on Dialogue in the New Millennium, –. Amsterdam: John Benjamins. van der Sandt, Rob A., and Emar Maier (). Denials in discourse. In Michigan Linguistics and Philosophy Workshop, University of Nijmegen. van der Wouden, Ton (). Negative Contexts. Doctoral Dissertation, University of Groningen, Groningen. van der Wouden, Ton (). Litotes and downward monotonicity. In H. Wansing (ed.), Negation: A Notion in Focus, –. Berlin: Walter de Gruyter. van der Wouden, Ton (). Negative Contexts: Collocation, Polarity and Multiple Negation. Amsterdam: Benjamins. van der Wouden, Ton (). Three modal verbs. In S. Watts, J. West, and H.-J. Solms (eds.), Zur Verbmorphologie germanischer Sprachen, –. Tubingen: Niemeyer. Van de Velde, Freek, and Muriel Norde (). Exaptation: Taking stock of a controversial notion in linguistics. In Muriel Norde and Freek Van de Velde (eds.), Exaptation and Language Change, –. Amsterdam: Benjamins. van Dijk, T. A., and W. Kintsch (). Strategies of Discourse Comprehension. New York: Academic Press. van Eijck, Jan (). Quantification (Quantoren). In Hugo Steger and Herbert Ernst Wiegand (eds.), Handbücher zur Sprach- und Kommunikations-wissenschaft, –. Berlin and New York: Walter de Gruyter. van Eijck, Jan (). Presuppositions and Information Updating. Centrum voor Wiskunde en Informatica. van Gelderen, Elly (). Grammaticalization as Economy. Berlin: John Benjamins. van Gelderen, Elly (). Negative cycles. Linguistic Typology  (): –. van Gelderen, Elly (). Cyclical change, an introduction. In Elly van Gelderen (ed.), Cyclical Change, –. Amsterdam: Benjamins. van Gelderen, Elly (). The Linguistic Cycle: Language Change and the Language Faculty. Oxford: Oxford University Press. van Gelderen, Elly (ed.) (). Cyclical Change Continued. Amsterdam: Benjamins. van Gelderen, Elly (). The negative existential and other cycles: Jespersen, Givón, and the Copula Cycle. Ms., Arizona State University. . Vanhaudenhuyse, Audrey, Quentin Noirhomme, Luaba J.-F. Tshibanda, Marie-Aurelie Bruno, Pierre Boveroux, Caroline Schnakers, Andrea Soddu, et al. (). Default network connectivity reflects the level of consciousness in non-communicative brain-damaged patients. Brain  (): –. DOI: ./brain/awp. van Herten, M., H. H. J. Kolk, and D. Chwilla (). An ERP study of P effects elicited by semantic anomalies. Cogn. Brain Res. : –. Van Kemenade, Ans (). Jespersen’s cycle revisited: Formal properties of grammaticalization. In Susan Pintzuk, George Tsoulas, and Anthony Warner (eds.), Diachronic Syntax: Models and Mechanisms, –. Oxford: Oxford University Press.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Vannuscorps, Gilles, and Alfonso Caramazza (). Typical action perception and interpretation without motor simulation. Proceedings of the National Academy of Sciences : –. Van Olmen, Daniel (). Review of Alexandra Y. Aikhenvald’s () Imperatives and Commands. Linguistic Typology : –. Van Olmen, Daniel (). Reproachatives and imperatives. Linguistics : –. van Petten, Cyma (). Words and sentences: Event-related brain potential measures. Psychophysiology : –. Vanrell, Maria Del Mar M., Ignasi Mascaró, Francesc Torres-Tamarit, and Pilar Prieto (). Intonation as an encoder of speaker certainty: Information and confirmation yes-no questions in Catalan. Language and Speech  (): –. van Rooy, Robert (). Negative polarity items in questions: Strength as relevance. Journal of Semantics : –. van Rooy, Robert, and Maria Šafářová (). On polar questions. In Robert B. Young and Yuping Zhou (eds.), Proceedings of Semantics and Linguistic Theory , –. Ithaca, NY: CLC Publications. Van Schaaik, Gerjan (). Turkish. In Kahrel and van den Berg (eds.), Typological Studies in Negation, –. van Son, Chantal, Emiel van Miltenburg, and Roser Morante (). Building a dictionary of affixal negations. Proceedings of the Workshop on Extra-Propositional Aspects of Meaning in Computational Linguistics, –. Osaka, Japan. Varela-Pérez, José Ramón (). Variation and Change in Negative Constructions in Contemporary British English: A Corpus-Based Approach. Ph.D. dissertation, University of Santiago de Compostela. Varvoutis, J., and M. Hackl (). Parsing quantifiers in object position. A talk delivered at the th CUNY Conference on Human Sentence Processing, CUNY, NY. Varzi, Achille C. (). The talk I was supposed to give. In A. Bottani and R. Davies (eds.), Modes of Existence: Papers in Ontology and Philosophical Logic, –. Frankfurt: Ontos Verlag. Varzi, Achille C. (). Omissions and causal explanations. In F. Castellani and J. Quitterer (eds.), Agency and Causation in the Human Sciences, –. Paderborn: Mentis. Vasishth, Shravan (). Monotonicity constraints on negative polarity in Hindi. In Mary M. Bradshaw, David Odden, and Derek Wyckoff (eds.), Ohio State Working Papers in Linguistics : –. Ohio: The Ohio State University. Vasishth, Shravan, Sven Brüssow, Richard L. Lewis, and Heiner Drenhaus (). Processing polarity: How the ungrammatical intrudes on the grammatical. Cognitive Science  (): –. Veloudis, Ioannis (). Negation in Modern Greek. Ph.D. dissertation, University of Reading. Vendler, Zeno (a). Each and every, any and all. In Zeno Vendler, Linguistics in Philosophy, –. Ithaca, NY: Cornell University Press. Vendler, Zeno (b). Linguistics in Philosophy. Ithaca, NY: Cornell University Press. Verhagen, Josje (). Temporal adverbials, negation and finiteness in Dutch as a second language: A scope-based account. IRAL, International Review of Applied Linguistics in Language Teaching  (): –. Vermeulen, Reiko (). On the position of topics in Japanese. The Linguistic Review  (): –. Veselinova, Ljuba (). Standard and special negators in the Slavonic languages: Synchrony and diachrony. In Björn Hansen and Jasmina Grković-Major (eds.), Diachronic Syntax of the Slavonic Languages, –. Vienna: Sagner. Veselinova, Ljuba (). Negative existentials: A cross-linguistic study. Italian Journal of Linguistics : –. Veselinova, Ljuba (). The negative existential cycle revisited. Linguistics  (): –. Veselinova, Ljuba (). The negative existential cycle viewed through the lens of comparative data. In van Gelderen (ed.), Cyclical Change Continued, –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Veselinova, Ljuba and Arja Hamari (eds.) (forthcoming). The Negative Existential Cycle from a Historical-Comparative Perspective. Berlin: Language Science Press. Vicente, Luis (). Short negative replies in Spanish. Linguistics in the Netherlands : –. Vining, Alexander Q., and Heidi L. Marsh (). Information seeking in capuchins (Cebus apella): A rudimentary form of metacognition? Animal Cognition : –. Visconti, Jacqueline (). From “textual” to “interpersonal”: On the diachrony of the Italian particle mica. Journal of Pragmatics : –. Vizitelly, Frank (). A Desk-Book of Errors in English. New York: Funk & Wagnalls. Vlach, Frank (). On situation semantics for perception. Synthese  (): –. Vlach, Frank (). Temporal adverbials, tenses and the perfect. Linguistics and Philosophy : –. Vogt, A., B. Kaup, and C. Dudschig (). When words are upside down: Language–space associations in children and adults. Journal of Experimental Child Psychology : –. Völter, Christoph. J., and Josep Call (). Causal and inferential reasoning in animals. In G. M. Burghardt, I. M. Pepperberg, C. T. Snowdon, and T. Zentall (eds.), APA Handbook of Comparative Psychology, Vol. : Perception, Learning, and Cognition, –. Washington, DC: American Psychological Association. Völter, Christoph J., Brandon Tinklenberg, Josep Call, and Amanda M. Seed (). Comparative psychometrics: Establishing what differs is central to understanding what evolves. Phil. Trans. R. Soc. B : . Volterra, V., and F. Antinucci (). Negation in child language: A pragmatic study. In E. Ochs and B. Schieffelin (eds.), Developmental Pragmatics, –. New York: Academic Press. von der Gabelentz, Georg () []. Die Sprachwissenshaft: Ihre Aufgaben, Methoden und bisherigen Ergebnisse. Leipzig: Weigel. (Reprint Tübingen: Narr .) Vonen, Arnfinn Muruvik (). Negation in Tokelauan. In Hovdhaugen and Mosel (eds.), Negation in Oceanic Languages, –. von Fintel, Kai (). Bare plurals, bare conditionals, and only. Journal of Semantics : –. von Fintel, Kai (). NPI licensing, Strawson entailment, and context dependency. Journal of Semantics  (): –. von Fintel, Kai (). Would you believe it? The King of France is back! (Presuppositions and truthvalue intuitions). In A. Bezuidenhout and M. Reimer (eds.), Descriptions and beyond: An Interdisciplinary Collection of Essays on Definite and Indefinite Descriptions and Other Related Phenomena. Oxford: Oxford University Press. von Fintel, Kai, and Sabine Iatridou (). Epistemic containment. Linguistic Inquiry : –. von Fintel, Kai, and Sabine Iatridou (). Anatomy of a modal construction. Linguistic Inquiry : –. von Fintel, Kai, and Lisa Matthewson (). Universals in semantics. The Linguistic Review  (–): –. von Stechow, Arnim (). Comparing semantic theories of comparison. Journal of Semantics : –. von Stechow, Arnim (). Current issues in the theory of focus. In von Stechieter and Wunderlich (eds.), Semantik: Ein Internationales Handbuch der Zeitgenosssischen Forschun. Berlin: Mouton de Gruyter. von Wright, G. H. (). On the logic of negation. Societas Scientiarum Fennica, Commentationes Physico-Mathematicae, –. Vossen, Frens (). On the Typology of the Jespersen Cycles. Ph.D. dissertation, University of Antwerp. Vossen, Frens, and Johan van der Auwera (). The Jespersen cycles seen from Austronesian. In Maj-Britt Mosegaard Hansen and Jacqueline Visconti (eds.), The Diachrony of Negation, –. Amsterdam: Benjamins. Vuillermet, Marine (). A Grammar of Ese Ejja, a Takanan Language of the Bolivian Amazon. Ph.D. dissertation, Université Lumière Lyon .

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Vulkanovic, Relja (). Development of negation in French: A quantitative model. Journal of Quantitative Linguistics  (–): –. Vygotsky, Lev S. (). Thought and Language. Cambridge, MA: MIT Press. Wagenaar, Kernelis (). Étude sur la négation en ancien espagnol jusqu’au XVe siècle. Groningen: Noordhoff. Wagers, M. W., E. F. Lau, and C. Phillips (). Agreement attraction in comprehension: Representations and processes. Journal of Memory and Language  (): –. Wagner, Michael (). Association by movement: Evidence from NPI-licensing. Natural Language Semantics : –. Wagner, Michael (). A compositional analysis of contrastive topics. In Muhammad Abdurrahman, Anisa Schardl, and Martin Walkow (eds.), Proceedings of NELS , Vol., –. Wagner, Michael (). Focus and givenness: A unified approach. In Ivona Kučerová and Ad Neeleman (eds.), Information Structure. Contrasts and Positions, –. Cambridge: Cambridge University Press. Wagner, Michael (). Contrastive topics decomposed. Semantics and Pragmatics : .–.. Waite, Jeffrey (). Negatives in Maori: A lexical-functional approach. Te Reo : –. Waldmann, Michael R., Martina Schmid, Jared Wong, and Aaron P. Blaisdell (). Rats distinguish between absence of events and lack of evidence in contingency learning. Animal Cognition : –. Wales, R., and R. Grieve (). What is so difficult about negation? Perception and Psychophysics : –. Walker, C. M., and A. Gopnik (). Toddlers infer higher-order relational principles in causal learning. Psychological Science  (): –. Walkow, Martin (). When to ask an inner negation polar question. In Maria Biezma and Jesse A. Harris (eds.), UMOP , –. Amherst: GLSA. Wallage, Phillip (). Negation in Early English: Parametric variation and grammatical competition. Ph.D. dissertation, University of York. Wallage, Phillip (). Jespersen’s Cycle in Middle English: Parametric variation and grammatical competition. Lingua : –. Wallage, Phillip (). Negative inversion, negative concord and sentential negation in the history of English. English Language and Linguistics  (): –. Wallage, Phillip (). Functional differentiation and grammatical competition in the English Jespersen Cycle. Journal of Historical Syntax : –. Wallage, Phillip (). Identifying the role of pragmatic activation in changes to the expression of English negation. In Pierre Larrivee and Chungmin Lee (eds.), Negation and Negative Polarity: Cognitive and Experimental Perspectives, –. Berlin: Springer International Publishing. Wallage, Phillip (). Identifying the role of pragmatic activation in changes to the expression of English negation. In Pierre Larrivée and Chungmin Lee (eds.), Negation and Polarity: Experimental Perspectives, –. Cham: Springer. Wallage, Phillip () Negation in Early English: Grammatical and Functional Change. Cambridge: Cambridge University Press. Wallenberg, J., and J. Fruehwald (). Optionality is stable variation is competing grammars. Talk given at the Formal Ways of Analyzing Variation Workshop, Reykjavik, Iceland. Wansing, Heinrich (). Proofs, disproofs, and their duals. In Lev Beklemishev, Valentin Goranko, and Valentin Shehtman (eds.), Advances in Modal Logic, : –. London: College Publications. Wason, P. C. (). The processing of positive and negative information. Quarterly Journal of Experimental Psychology  (): –. Wason, P. C. (). Response to affirmative and negative binary statements. British Journal of Psychology : –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Wason, P. C. (). The contexts of plausible denial. Journal of Verbal Learning and Verbal Behavior  (): –. Wason, P. C. (). In real life negatives are false. Logique et analyse : –. Wason, P. C., and P. N. Johnson-Laird (). Psychology of Reasoning: Structure and Content. Boston: Harvard University Press. Wason, P. C., and S. Jones (). Negatives: Denotation and connotation. British Journal of Psychology : –. Wason, P. C., and S. S. Reich (). A verbal illusion. Quarterly Journal of Experimental Psychology  (): –. DOI: ./. Watanabe, Akira (). The genesis of negative concord: Syntax and morphology of negative doubling. Linguistic Inquiry  (): –. Watson, J. M. (). Referential description by children in negative form. British Journal of Psychology : –. Wegner, D. M. (). Ironic processes of mental control. Psychological Review : –. Wegner, D. M., M. Ansfield, and D. Pilloff (). The putt and the pendulum: Ironic effects of the mental control of action. Psychological Science : –. Wegner, D. M., J. W. Shortt, A. W. Blake, and M. S. Page (). The suppression of exciting thoughts. Journal of Personality and Social Psychology : –. Weir, Andrew (). Fragments and Clausal Ellipsis. University of Massachusetts Amherst Ph.D. dissertation. Weir, E. M. Helen (). Nadëb. In Kahrel and van den Berg (eds.), Typological Studies in Negation, –. Weiß, Helmut (). Three types of negation: A case study in Bavarian. In Sjef Barbiers, Leonie Cornips, and Susanne van der Kleij (eds.), Syntactic Microvariation, –. Meertens Institute Electronic Publications in Linguistics. . Wellwood, A., R. Pancheva, V. Hacquard, and C. Phillips (). The anatomy of a comparative illusion. Journal of Semantics : –. Wester, Ruth (). Negation in Awyu-Dumut Languages. Ph.D. dissertation, Free University of Amsterdam. White, Lydia (/). The verb-movement parameter in second language acquisition. Language Acquisition : –. White, Lydia (). Adverb placement in second language acquisition: Some effects of positive and negative evidence in the classroom. Second Language Research : –. White, Lydia (a). Long and short verb movement in second language acquisition. Canadian Journal of Linguistics : –. White, Lydia (b). On triggering data in L acquisition: A reply to Schwartz and Gubula-Ryzak. Second Language Research : –. White, Lydia (). Second Language Acquisition and Universal Grammar. Cambridge: Cambridge University Press. White, Lydia (). Formal linguistics and second language acquisition. In D. Miller, F. Bayram, J. Rothman, and L. Serratrice (eds.), Bilingual Cognition and Language: The State of the Science Across its Subfields, –. Amsterdam: John Benjamins. Whiten, Andrew (). Humans are not alone in computing how others see the world. Animal Behaviour : –. Whiten, Andrew, and Richard W. Byrne (). Tactical deception in primates. Behavioral and Brain Sciences : –. Wible, David, and Eva Chen (). Linguistic limits on metalinguistic negation: Evidence from Mandarin and English. Language and Linguistics  (): –. Wilbur, Ronnie B. (). Phonological and prosodic layering of nonmanuals in American Sign Language. In Karen Emmorey and Harlan Lane (eds.), The Signs of Language Revisited, –. Mahwah, NJ: Lawrence Erlbaum.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Wilcox, Sherman (). Gesture and language: Cross-linguistic and historical data from signed languages. Gesture : –. Wilcox, Sherman (). Routes from gesture to language. In Elena Pizzuto, Paola Pietrandrea, and Raffaele Simone (eds.), Verbal and Signed Languages: Comparing Structures, Constructs, and Methodologies, –. Berlin: Mouton de Gruyter. Wilkinson, Karina (). The scope of even. Natural Language Semantics : –. Williams, Alexander (). Arguments in Syntax and Semantics. Key Topics in Syntax. Cambridge: Cambridge University Press. Williams, C. J. F. (). Saint Anselm and his biographers. Downside Review : –. Willis, David (). Motivating the emergence of new markers of sentential negation: The case of Welsh ddim. Diachronica : –. Willis, David (a). A Minimalist approach to Jespersen’s cycle in Welsh. In Dianne Jonas, John Whitman, and Andrew Garrett (eds.), Grammatical Change: Origins, Natures, Outcomes, –. Oxford: Oxford University Press. Willis, David (b). Negative polarity and the Quantifier Cycle: Comparative diachronic perspectives from European languages. In Pierre Larrivée and Richard Ingham (eds.), The Evolution of Negation: Beyond the Jespersen Cycle, –. Berlin: Mouton de Gruyter. Willis, David (a). Negation in the history of the Brythonic Celtic languages. In David Willis, Christopher Lucas, and Anne Breitbarth (eds.), The History of Negation in the Languages of Europe and the Mediterranean, Vol. : Case Studies, –. Oxford: Oxford University Press. Willis, David (b). Negation in the history of the Slavonic languages. In David Willis, Christopher Lucas, and Anne Breitbarth (eds.), The History of Negation in the Languages of Europe and the Mediterranean, Vol. : Case Studies, –. Oxford: Oxford University Press. Willis, David, Christopher Lucas, and Anne Breitbarth (a). Comparing diachronies of negation. In David Willis, Christopher Lucas, and Anne Breitbarth (eds.), The History of Negation in the Languages of Europe and the Mediterranean, Vol. : Case Studies, –. Oxford: Oxford University Press. Willis, David, Christopher Lucas, and Anne Breitbarth (eds.) (b). The History of Negation in the Languages of Europe and the Mediterranean, Vol. : Case Studies. Oxford: Oxford University Press. Willmott, Jo (). Negation in the history of Greek. In David Willis, Christopher Lucas, and Anne Breitbarth (eds.), The History of Negation in the Languages of Europe and the Mediterranean. Vol. : Case Studies, –. Oxford: Oxford University Press. Wilson, Deirdre (a). Presupposition and Non-Truth-Conditional Semantics. New York: Academic Press. Wilson, Deirdre (b). Presupposition, assertion, and lexical items. Linguistic Inquiry  (): –. Wilson, Deirdre (). Metarepresentation in linguistic communication. In Dan Sperber (ed.), Metarepresentations: A Multidisciplinary Perspective, –. Oxford: Oxford University Press. Wilson, Deirdre (). The conceptual-procedural distinction: Past, present and future. In Victoria Escandell-Vidal, Manuel Leonetti, and Aiofe Ahern (eds.), Procedural Meaning: Problems and Perspectives, –. Bingley: Emerald Group. Wilson, Deirdre (). Reassessing the conceptual-procedural distinction. Lingua –: –. Wilson, Deirdre, and Dan Sperber (). Mood and the analysis of non-declarative sentences. In Jonathan Dancy, J. Michael Moravcsik, and Charles Taylor (eds.), Human Agency: Language, Duty and Value, –. Stanford, CA: Stanford University Press. Wilson, Deirdre, and Dan Sperber (). Meaning and Relevance. Cambridge: Cambridge University Press. Winters, Margaret E. (). Innovations in French negation: A cognitive grammar account. Diachronica  (): –. Wode, Henning (). Four early stages in the development of L negation. Journal of Child Language  (): –. Wolfram, Walt, and Donna Christian (). Appalachian Speech. Arlington, VA: Center for Applied Linguistics.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Wolfram, Walt, and Ralph Fasold (). The Study of Social Dialects in American English. Englewood Cliffs, NJ: Prentice Hall. Wood, Sandra (). Semantic and Syntactic Aspects of Negation in ASL. MA thesis, Purdue University. Woolford, Ellen (). Lexical Case, inherent Case, and argument structure. Linguistic Inquiry : –. Wurm, Moritz F., and Alfonso Caramazza (). Distinct roles of temporal and frontoparietal cortex in representing actions across vision and language. Nature Communications : . Wynn, K. (). Addition and subtraction by human infants. Nature  (): . Wynn, K., and W. C. Chiang (). Limits to infants’ knowledge of objects: The case of magical appearance. Psychological Science  (): –. Xherija, Orest (). Weak and strong NPIs: ‘Nobody’ and ‘anybody’ in Albanian and Modern Greek. Proceedings of Texas Linguistic Society. Xiang, M., B. Dillon, and C. Phillips (). Illusory licensing effects across dependency types: ERP evidence. Brain and Language  (): –. Xiang, M., J. Grove, and A. Giannakidou (). Dependency-dependent interference: NPI interference, agreement attraction, and global pragmatic inferences. Frontiers in Psychology : . Xiang, M., J. Grove, and A. Giannakidou (). Semantic and pragmatic processes in the comprehension of negation: An event related potential study of negative polarity sensitivity. Journal of Neurolinguistics : –. Xiang, Yimei (), ONLY: An NPI-licenser and NPI-unlicenser. Journal of Semantics : –. Xu, Ming (). Causation in branching time (I): Transitions, events and causes. Synthese : –. Yaeger-Dror, Malcah (). Intonational prominence on negatives in English. Language and Speech : –. Yaeger-Dror, Malcah (). Register and prosodic variation: A cross language comparison. Journal of Pragmatics : –. Yang, Charles (). The Price of Linguistic Productivity: How Children Learn to Break the Rules of Language. Cambridge, MA: MIT Press. Yang, Jun Hui, and Susan D. Fischer (). Expressing negation in Chinese Sign Language. Sign Language & Linguistics  (): –. Yanilmaz, Aydogan, and John E. Drury (). Prospective NPI licensing and intrusion in Turkish. Language, Cognition and Neuroscience  (): –. Yanwei Jin and Jean-Pierre Koenig (). A cross-linguistic study of expletive negation. CSSP Paris, November –, . Yoon, Suwon (). Two types of negative indefinite in Korean. In John Whitman, Ponghyung Lee, Chang-Kook Suh, Sungdai Cho, Yeun-Jin Jung, Jae Young Chung, and Jeong-Im Han (eds.), Inquiries into Korean Linguistics III, –. Seoul: Hankook Publishing Co. Yoon, Suwon (). ‘Not’ in the Mood: The Syntax, Semantics, and Pragmatics of Evaluative Negation. Ph.D. dissertation, University of Chicago. Yoon, Youngeun (). Total and partial predicates and the weak and strong interpretations. Natural Language Semantics  (): –. Yoshimura, Akiko (). A cognitive constraint on negative polarity phenomena. In Berkeley Linguistic Society, pp. –. Berkeley. Yoshimura, Keiko (). Scope theory vs. polarity theory: The Japanese focus particle -sae. Proceedings from the Annual Meeting of the Chicago Linguistic Society, –. Chicago: CLS. Young, Robert, and William Morgan (). The Navajo Language. Albuquerque, NM: University of New Mexico Press. Yun, Jiwon, So Young Lee, and John Drury (). Negative polarity illusion in Korean. The Proceedings of WAFL .

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Yurchenko, Anna, Dirk-Bart den Ouden, Jack Hoeksema, Olga Dragoy, John C. J. Hoeks, and Laurie A. Stowe (). Processing polarity: ERP evidence for differences between positive and negative polarity. Neuropsychologia  (): –. Zaccarella, Emiliano, Lars Meyer, Michiru Makuuchi, and Angela D. Friederici (). Building by syntax: The neural basis of minimal linguistic structures. Cerebral Cortex  (): –. Zagona, Karen (). The Syntax of Spanish. Cambridge: Cambridge University Press. Zanuttini, Raffaella (). Syntactic Properties of Sentential Negation: A Comparative Study of Romance Languages. Ph.D. dissertation, University of Pennsylvania. Zanuttini, Raffaella (). Negation and Clausal Structure: A Comparative Study of Romance Languages. Oxford: Oxford University Press. Zanuttini, Raffaella (). Sentential negation. In Mark Baltin and Chris Collins (eds.), The Handbook of Contemporary Syntactic Theory, –. Oxford: Blackwell. Zanuttini, Raffaella, and Paul Portner (). The characterization of exclamatives clauses in Paduan. Language : –. Zeevat, Henk (). Presupposition and accommodation in update semantics. Journal of Semantics  (): –. Zeijlstra, Hedde (a). Sentential Negation and Negative Concord. Ph.D. dissertation, University Amsterdam. Utrecht: LOT. Zeijlstra, Hedde (b). Syntactic vs. semantic negation. In Cécile Meier and Matthias Weisgerber (eds.), Proceedings of the Conference Sinn und Bedeutung, vol.  Arbeitspapiere / Fachbereich Sprachwissenschaft. Konstanz: Universität Konstanz. Zeijlstra, Hedde (). Negation in natural language: On the form and meaning of negative elements. Language and Linguistics Compass : –. Zeijlstra, Hedde (a). Negative concord is syntactic agreement. Ms., University of Amsterdam. . Zeijlstra, Hedde (b). On the syntactic flexibility of formal features. In Theresa Biberauer (ed.), The Limits of Syntactic Variation, –. Amsterdam: Benjamins. Zeijlstra, Hedde (a). Emphatic multiple negative expressions in Dutch. The Linguistic Review  (): –. Zeijlstra, Hedde (b). On French negation. In I. Kwon, H. Pritchett, and J. Spence (eds.), Proceedings of the th Annual Meeting of the Berkeley Linguistics Society. Berkeley: BLS. Zeijlstra, Hedde (). On the syntactically complex status of negative indefinites. Journal of Comparative Germanic Linguistics : –. Zeijlstra, Hedde (). Negation and polarity. In M. Den Dikken (ed.), The Cambridge Handbook of Generative Syntax, –. Cambridge: Cambridge University Press. Zeijlstra, Hedde (). On the uninterpretability of interpretable features. In Peter Kosta, Steven L. Franks, Teodora Radeva-Bork, and Lilia Schürcks (eds.), Minimalism and beyond: Radicalizing the Interfaces, –. Amsterdam: Benjamins. Zeijlstra, Hedde (). The morpho-syntactic realisation of negation. In Tibor Kiss and Artemis Alexiadou (eds.), Syntax—Theory and Analysis, –. HSK  Vol. . Berlin: Mouton de Gruyter. Zeijlstra, Hedde (). Diachronic developments in the domain of negation. Language and Linguistics Compass  (): –. Zeijlstra, Hedde (a). Does NEG-Raising involve NEG-Raising? Topoi: An International Review of Philosophy : –. Zeijlstra, Hedde (b). Universal Quantifier PPIs. Glossa  (): .–. Zeijlstra, Hedde (a). Does Neg-raising involve Neg-raising? Topoi : –. Zeijlstra, Hedde (b). Negation and Negative Indefinites. To appear with Oxford University Press. Zentall, T. R., D. M. Andrews, and J. P. Case (). Sameness may be a natural concept that does not require learning. Psychological Science  (): –.

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi





Zentall, T. R., C. A. Edwards, B. S. Moore, and D. E. Hogan (). Identity: The basis for both matching and oddity learning in pigeons. Journal of Experimental Psychology: Animal Behavior Processes  (): . Zeshan, Ulrike (). Hand, head, and face: Negative Constructions in sign languages. Linguistic Typology  (): –. Zeshan, Ulrike (ed.) (a). Interrogative and Negative Constructions in Sign Languages. Nijmegen: Ishara Press. Zeshan, Ulrike (b). Negative and interrogative constructions in sign languages: A case study in sign language typology. In Ulrike Zeshan (ed.), Interrogative and Negative Constructions in Sign Languages, –. Nijmegen: Ishara Press. Zeshan, Ulrike (c). Negative and interrogative structures in Turkish Sign Language (TİD). In Ulrike Zeshan (ed.), Interrogative and Negative Constructions in Sign Languages, –. Nijmegen: Ishara Press. Zeshan, Ulrike (). Irregular negatives in sign languages. In Matthew S. Dryer and Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at ) Zhang, Linmin, and Liina Pylkkänen (). Semantic composition of sentences word by word: MEG evidence for shared processing of conceptual and logical elements. Neuropsychologia : –. Zhou, Peng, Stephen Crain, and Rosalind Thornton (). Children’s knowledge of double negative structures in Mandarin Chinese. Journal of East Asian Linguistics : –. Ziegeler, Debra (). On the empty O-corner of the Aristotelian Square: A view from Singapore English. Journal of Pragmatics : –. Zimmer, Karl (). Affixal negation in English and other languages. Supplement to Word, Monograph . Zimmermann, Malte (). Contrastive focus and emphasis. Acta Linguistica Hungarica : –. Zimmermann, Thomas Ede (). On the proper treatment of opacity in certain verbs. Natural Language Semantics : –. Zucchi, Sandro (). Events and situations. Annual Review of Linguistics : –. Zufferey, Sandrine, Jacques Moeschler, and Anne Reboul (). Implicatures. Cambridge: Cambridge University Press. Zwaan, R. A., and C. J. Madden (). Embodied sentence comprehension. In D. Pecher and R. A. Zwaan (eds.), The Grounding of Cognition: The Role of Perception and Action in Memory, Language, and Thinking, –. Cambridge: Cambridge University Press. Zwaan, R. A., R. A. Stanfield, and R. H. Yaxley (). Language comprehenders mentally represent the shapes of objects. Psychological Science : –. Zwarts, Frans (). Negatief polair uitdrukkingen. GLOT  (): –. Zwarts, Frans (). Categoriale grammatica en algebraïsche semantiek: Een studie naar negatie en polariteit in het Nederlands? Ph.D. thesis, University of Groningen. Zwarts, Frans (). Nonveridical contexts. Linguistic Analysis : –. Zwarts, Frans (). A hierarchy of negative expressions. In Heinrich Wansing (ed.), Negation: A Notion in Focus, –. Berlin and New York: de Gruyter. Zwarts, Frans (). Three types of polarity. In Fritz Hamm and Erhard Hinrichs (eds.), Plurality and Quantification, Vol. , –. Dordrecht: Kluwer. Zwicky, Arnold M. (). In a manner of speaking. Linguistic Inquiry  (): –. Zwicky, Arnold M., and Jerry Sadock (). Ambiguity tests and how to fail them. In J. Kimball (ed.), Syntax and Semantics : –. New York: Academic Press.

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

I

...................

A A-not-A  ABA – Abstraction –, –,  accentuation – acceptability speech judgement task  accommodation –, , –, ,  in ellipsis  Accord Maximization Principle  acquisition adults –, – children , , , –, –, – activation –, , –, , , , –, –,  brain – neuronal  adverb –, –, –, –, –, , , , , , , , , –, –, –, , , – of quantification –, ,  adverbial  emphasizer ,  modification  affixal ordering  affixation –, – Afro-Asiatic  Afrikaans  agrammatism – Agree , , , , –, –, ,  multiple Agree –, , –,  agreement marking –,  algorithmic – already , , , –, , ,  alternatives –, ,  alternative semantics , –, –,  computation of alternatives –, – ambiguity , –, –, , – ambiguist n. Amharic  ancora  Andoke  animal cognition – answer , –, , , ,  answer particle ,  anti-additive , , –, , –, – anti-morphic –, –, –

anti-multiplicative  antisymmetrical approach  anti-veridicality –, , , ,  antonyms –, , , , –,  any , , –, , –, , –, –, , –, –, –, –, , , –, –, ,  anymore , – anyone  anything –, , , – aphasia , – Arabic  Egyptian , ,  Syrian  argument cycle  argumental reinforcer Aristotelian square of oppositions see square of oppositions Arizona Tewa  aspect –, , –, , , , –, , –,  assertion –, , , – assertability –, – operator (Bochvar) –,  associative learning ,  at-issueness – Ateso  automaticity – Australian languages – Austronesian languages , – auxiliary verb –, , –, , –, –, , , ,  initial  Avava  Awju-Ok 

B

Bafut – Bahnaric languages –,  Bantawa , ,  Barambu  Bari  Basque , – be surprised  Beck effect , –, – Berber , ,  Biak –

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi





bias – contextual evidence bias  original speaker bias  bilectalism  bilingual(ism) , , –, , –, , –, –, ,  binding , ,  blocking ,  brain , –, –, –, , , –, – brauchen , – Broca’s area –

C

c-command , , , , , , –, , , , , ,  C features ,  Cancellation –, – canonical logical form of negation , , ,  Cape-Verdean Creole  Carib  Case –,  ablative  dative  ergative –,  nominative pre-verbal subject  nominative/absolutive  nominative/accusative – Catalan –, , , , –, –, –, , –, , , , , , , –, –, – categorical judgement ,  categorization of affixal negation , –, – causal event  reasoning  relation  report – causation , – Chadic  Chamic languages – Chamorro  change in grammatical category  choice function  Classneg-marker –, – clitic –, –, –, , , –, , , , –,  placement –, –, ,  pronoun , –, –, –, ,  code-mixing  code-switching  common ground –, ,  comparatives , –, –, –, , – complementizer , , , , , , , ,  Complementizer Phrase –, –, –, , , – CP specifier , 

complexity , , , – compounding  computational systems  communication , –, – conditional , ,  conflict ,  conjunction  connegative ,  constituent ordering , –, , , , , ,  canonical , , , ,  ‘free word order’ language  non canonical ,  variant constituent ordering  context , –, , –, –, , , , ,  contextual alleviation –,  continuation/rectification –, – continuity  contraction ,  contradiction contour –,  conventional implicature –, –, , , – conversion  co-occurrence ,  copula inversion  Cornish ,  cortical silent period – count/mass –, , – Croft’s cycle , – see also negative existential cycle cyclical development –, – Czech –, , 

D

Danish , , ,  De Morgan’s laws –, – definite description , , –, –, –, , – of events , –, –,  degree , –, –, , – modifier  denial –, –, , , , , –, –, , , , –, , – see also predicate denial dependency –, , , –,  syntactic , – descriptive anaphora –, – descriptive event anaphora –,  Determiner Phrase ,  development , ,  diachronic cline , –, , ,  dialect –, –, , , – Dida – directionality , – discourse –, –, ,  anaphora 

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 discourse-old information – particles  Discourse Representation Theory  disjunction under negation , , ,  dispositional affect  distributive operator ,  distributivity – Dongo – donkey anaphora – double agreement  double negation see negation, double doubt – downward entailment , , –, –, , , ,  downward entailing operators –, – Strawson downward entailment –, –, – Dutch , , , , –, , , –, , , –, –, –, , –, , , , , ,  Belgian ,  Katwijk  Middle – Old – Southern  Dravidian ,  Dynamic Semantics –

E

each  East Futunan – Eastern Nilotic languages – Eastern Oromo  Economy – een pretje  elasticity  electroencephalography –, –, –, , , , – mu rhythms  theta rhythms – electrophysiology –, –,  ellipsis , –, –, , –,  A-movement in , ,  identity in –,  structure in ,  else  embedded clause see subordinate clause embodiment – emotional mood  emphasis , , , , –, , –, , , –, , –, ,  emphatic affirmation/affirmative , – emphatic particle ,  English , , , , , –, , , , , –, , –, , –, –, , –, –, –, –, , –, , –, , –, , –, –, , –, , , , –



African American –, – Appalachian , –, , – Belfast –,  Buckie  Early Modern –, , , ,  Middle , , –, –, , ,  Old , , , , , ,  West Texas  White Alabama  enig ,  entailment –, –, , –, , , –, , –, , , –, –, , , – equative construction  equative negative construction  error rate – Ese Ejja  Estonian , – even , – Evenki , ,  event  quantification –, –, , ,  segmentation –, –,  semantics –, , , –, , , – time ,  event-related brain potential , , –, –, , , , –, –, , – eventuality ,  ever , , –,  every , – evolution , ,  Ewe –,  excluded middle –, –, , , –, , , , , –,  exhaustification – existential , –, , – determiner ,  negative , – quantification , , , ,  exponence , –, ,  expressive , –, , , ,  Extended Term Logic – extended Zwart’s hierarchy ,  extension , ,  eye tracking , 

F

factive predicate , –,  facts –, , – Fe’fe’  feature , , – clause-typing feature  [iNeg] = interpretable negative formal feature , –, –, –, –, , , –,  [Neg] = semantic negative feature –, –, 

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi





feature (cont.) [uNeg] = uninterpretable negative formal feature , –, , –, –, –, , –, –,  [uPol] = uninterpretable polarity formal feature –,  checking  economy – engineering  felicity conditions for negation , – few , ,  finite clause  Finnish , , ,  Flexibility ,  Focneg-marker –, , , –, ,  Focus –, , , –, –, –, ,  focus , , , , –, –, , , –, –, , , –, , , , – particle –, , , –, ,  semantics –, , – focus-sensitive particles  exclusive – scalar –, –, , – focused constituent , , –, , – formal renewal  fragments , –, , ,  framed events e –,  framing events E –, , –,  free choice item , – French , , –, , , , –, –, , , , , –, –, , , , , , , –, –, , –,  Belgian  Middle , – Montréal  Quebec  Old , , –,  Swiss ,  frequency of usage – Full Transfer Full Access Hypothesis ,  functional category/head/layer/position –, – functional magnetic resonance imaging , –, , –, 

G

garden-path effect  generalizers , – generic  Georgian  German , , , –, , –, –, , –, , , , , , –, , , –, –, –, , , , , , –, –, , , –, , –, –,  Middle High 

Middle Low  Old High , – gesture , , –, , , – Gιsιɖa Anii  give a red cent ,  Gödel’s Slingshot , , – Go-Nogo task – grammatical competition , –, ,  grammaticalization , –, –, –, ,  Grebo  Greek –, , , –, , , –, , , , –, , , , , , ,  Ancient , , , ,  Gude 

H

hardly , – Hausa  Hawai’ian  head movement –, ,  head negation see negation, head heritage , – heuristic – Hindi , –, ,  Hixkaryana  hoeven ,  homogeneity condition , – homogeneous eventuality  predicate – Hungarian , , , , , , 

I

idioms –, –, ,  idiomatic expressions , , –, – Igbo  illusions , –,  illusory licensing , – imperfective ,  imperfective aspect  impersonal verb  implication reversal  incompletive construction  indefinite –, , –, –, –, – individual differences , , – individual reasoning  inference , , –, , –,  by exclusion , –,  inferiority  inflationary effects , –,  inflected verb , –, , –,  information ,  focus  seeking – structure –, ,  inhibition –, –, , , –

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 instrument of causation  insufficiency ,  intended answer intension –,  Interface Hypothesis , –, –,  interlanguage – interval of degrees ,  intervener –, –,  intervention effect –, , –,  intonation , , –, , , , –, , , – intrusive licensing , – inverse scope , , ,  Irish –,  ironic effects ,  irrealis , , –,  Italian –, –, , –, , , –, , , , –, , , –, ,  Old , , , 

J

Jamaican Creole – Japanese , –, –, , , , , –, –, – Jarai  jemals ,  Jespersen’s cycle –, , , , , –, , , , , , , –, –, , , , , –, –,  judgment ,  qualitative , ,  quantitative ,  thetic 

K

Kannada , – Kanincin  Kayabai  Kanyok ,  Karitiána  Kiranti languages ,  Klao  Klima tests –, , –, , , , ,  Korean , –, –, –, , –, , –, , , , , –, ,  Kresh – Kru – Kwaa 

L

L , – L –, – Ladd’s ambiguity –, – language contact  Latin , , –, , , –, –, –, ,  lattice 



Law of Excluded Middle –, , ,  Law of Non-Contradiction –, ,  left-periphery , , –, –,  Leggbó  lexical categories , –, – lexicalization –,  lexical-semantic processing/network –,  Lewo ,  licensing , –, –, , –, –, , , –, , , –, , – context – operator –, , , – power – lift a finger , –, , , , – linguistic variety , –, – litotes , ,  local event , ,  locality , , –, ,  localized definite description , –, , ,  localized plural abstraction –,  locative/temporal deictics –, , ,  Logbara  logic, negation in –, –,  logic of negation in perceptual reports  logic of perception  logistic model of morphosyntactic change –,  Lopit –

M

Maa  Ma’di  main clause phenomena ,  Majang  Malay/Indonesian – Mam – Mandan  Mandarin Chinese , – Mandobo ,  manner , –, ,  Māori , – markedness , , , , – Marquesan  Masakin  matrix clause , , – matrix subject position – maximality – maximizer , , –, , –, –, – Mbili – Mbosi  meaning completeness , ,  Me’en  Metacognition  metalinguistic negation marker , –, – see also negation, metalinguistic idiomatic  internal 

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi





metalinguistic negation marker (cont.) external  unambiguous , –, –, – micro-syntactic variation ,  Minimal Trees Hypothesis  minimizer , , –, –, , –, –, –, –, –, , –, –,  vulgar , – mixed-effect regression modeling – modal –, , , –, ,  obviation , , ,  monadic event concepts – monoguist n. monotone decrease –, – mood , , , , ,  Mordvin  Moru-Ma’di languages – motor excitability –,  motor inhibition , – motor-action language –, –,  motor-evoked potentials –, – movement , , –, –, , , – see also head movement and syntactic movement multilingual –, ,  multiple-pair interpretation – Murle  Mursi 

N

N , , –,  Nadëb –,  Nall-problem , – nanosyntax ,  Nanti  natural language processing – Navajo  Nasioi  need , , , , – negated NP  negated sentences –, , , – negation –, –, –, , , , , , –, –, – affixal , , –, –, –, , –, –, , –, , , ,  anaphoric  annihilative  ascriptive , – asymmetric –, , ,  auxiliary , , –, , –, , , – binary – bipartite , –, , , ,  clause-internal ,  contradictory –, –, , –, , ,  contrary –, , –, , , , , –, , ,  contrastive/corrective , , –, –

constituent –, –, –, , –, –, , , , , , , , ,  in fragments –, – descriptive –, –, , –, , , , ,  direct , – discontinuous –,  double , –, –, , , –, , –, –, –, –, –, –, , –, , – double negation languages , , –, –, , – emphatic , , –, , , –, , , –, , ,  existential , , – external –, , –, , –, –, , –, –, , – formulaic external  final –, , , , –,  formulaic – head , , –, , –, , , –, – implicit –, ,  incorporated –, , , ,  indirect , – initial – internal , , –,  intervening , ,  metalinguistic –, –, –, , , , –, –, –, , , –, , – see also metalinguistic negation marker morphological , , –, , , –, ,  multiple , , , , , –, , –, , , – nexal –,  ordinary , –, , ,  phrasal –, , ,  preservative  privative , , , ,  prohibitive , –, , –,  propositional –, , , –, , , , – quadruple  quintuple  sentence/sentential –, –, –, –, –, –, , , , , , –, –, , , –, –, , , , , –, , , , , – single , , , , , , , –, , –, , –,  special –, ,  squatitive ,  standard –, –, –, , , – syntactic –, , , –, –, , – term –, –, –, –, – triple , , –, 

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 trivalent – types of negation , , , , , , – negation of noun , – negation of adjective – negation of verb –,  Negation Phrase –, , , –, , , , , –,  Big Negation Phrase Hypothesis ,  negative absorption  adverb , –, , , , –, –, , , – affix , , –, –, , , , , ,  auxiliary , –, , –, , –, , , – causatives – cleft , , ,  complementizer –,  concord , –, , –, , , –, , , –, , , , , –, –, –, , –, –, , , , –, – negative concord items , , –, , , –, –, , –, , , , , – strict negative concord – non-strict negative concord – connectives  constituents , , –, ,  cycle , ,  dependency , , – description doubling  events –, –, , –,  event description –, , –, , , ,  existential cycle –,  factorization  feature –, –, , –, , –, –, –, ,  see also feature sensitivity feature  head –, –, , –, , , , – imperative ,  imperfective  indefinites , , , –, –, –, –, , , ,  inversion –, –, , , ,  island –, –, , – licensing , – marker , –, –, , , , , , –, , , –,  acategorial status  operator , , –, –, , , , , –,  parentheticals , , – particle , , –, , , , , , , , , , 



perfective  position , –, , –, ,  quantification –,  quantifiers –, , , , –, –, , , , , –, , , –, –, , , –,  and object shift  and sentential scope , , –, ,  question see question, negative spread , –,  verb , –, , , , –, ,  Neg-first –,  Neg-first Principle –, ,  NEG-Q reading , –, – NEG-V reading , –, – Neg-raising –, –, , – neo-Davidsonian decomposition –, ,  separation –, ,  neuroimaging , –, , ,  neutral –, , ,  ni-minimizer , , , – ni-maximizer , –,  nicht mehr  niemand , , , ,  Niger-Congo languages , , –, ,  Niuean – no-negation – and the quantifier cycle –, – discourse functions of –, – frequency effects upon –, , – noch  nocht nicht  nominal predicate  nominalized clause – nominalized predicate  non-finite clause  non-matrix clause ,  non-pronominal object  non-standard –,  negation –, , –, , –, –,  varieties , , , , , –, – non-standardized – non-truth-conditional meaning , –, – non-verbal constructions  non-veridicality –, , , , , –, –, , , ,  nooit  not –, , , –, , , , , , , –, , –, , , –, –, –, , , – not-negation and scope –,  and the Jespersen’s cycle – discourse functions of – frequency effects upon  grammatical constraints upon –

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi





not-negation (cont.) historical persistence of – lexical diffusion of – nothing , – noughtiness ,  noughtly –, , , , , , , ,  Norwegian N-word , , , –, –, , , , 

O

object permanence ,  object preposing  object shift  objection –, –,  objects of perception , –, –, – omission ,  Oneida  only –, – ontogenesis of negation ooit  ook maar iets , – opposition , –, – Otuho  overhovedet 

P

P , –,  Paduan –,  Pana – paradigm –, , , –, ,  parsing , –,  Papuan languages  pas  pejorative , – perceptual and causal reports , –, ,  consistency –, –, , , ,  veridicality –, –, , , ,  persistive negative  person agreement –,  perspective taking PF movement  Piedmontese ,  pitch – plural definite description , –, ,  plural description –, , ,  plural event description , , ,  polarity –, –, –, – emphatic polarity ,  negative polarity –, , –, –, , , –,  adverb , ,  items see polarity items, negative sensitivity , –, –, –, , – sentential negative polarity , , –, –, , –

polarity items , –, , , –, , , , , –, –, –, –, –,  affective , –, –,  negative –, , –, –, –, –, , –, –, –, , , , –, , , –, , , –, , , , , , –, –, –, , , –, , –, , –, –, , –,  positive –, –, , –, , , , , , –, , , –,  strong negative –, , –, , , –, –, , , , ,  strong positive  superstrong negative , –, –, , –,  superstrong positive –, , ,  superweak –, , ,  superweak positive –, , , – weak – weak negative , –, ,  weak positive , – Polarity Phrase , ,  Polish , ,  politeness –,  Polynesian languages , –, – Portuguese , –, –, ,  Brazilian , , , –, – European , – potential licensor , – pragmatic activation –, – pragmatic conditions , , , , –, ,  pragmatic scale  pragmatic strengthening –, , , –, –, , , –, , ,  pragmatics –, , –, –, , –, , –, –, – predicate adversative  collective ,  negative –, –, –, – negative existential  scalar , – stage-level  predicate denial , , –, –, –, , – predicate term negation –, –, –, – prefixation , – premotor cortex –, , – prescriptive  presupposition , , , –, , –, , , –, –, , , , –, –, , –, –,  presuppositional negative marker , , , – priming , , , 

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 privative , , , ,  processing , , – sentence processing , –, –,  projection – propositional thought , ,  propositions –, , , –, , ,  prosody , , –, –, – proto-negation – proximate clitic  Pukapukan – pull-chain and push-chain ,  puzzles , , , 

Q

Qneg-marker –, – quantifier , , –, , –, –, , , , –,  see also negative quantifier cycle –, – raising –, , , ,  unselective quantifiers  question , , , , –, –, –, , –, , –, –, , ,  high negation –, – low negation  multiple wh-question –,  negative –, –, – positive – question semantics  rhetorical , –,  tag –, 

R rates of usage see frequency of usage realis ,  reanalysis –, –,  reconstruction – reduplication  referentiality , – referentially deficient expression  rejection –, –, , – removal ,  Rengao  ‘reordering’ languages  representation , –, –, , –, – response time , , , , , ,  restricted event quantifiers , , , ,  reversal , , –, –, –,  scale reversal  reversibility  Romance languages , , , , , , , –, , , , –, –, , –, , , , , , –, –, –, – Romanian –, , ,  Romeyka  Russian 

S



sample –, , – Sandonato  sarcasm  scalar , –, –, ,  implicature , , –, –, , ,  reinforcer , , – scale –, ,  schon  scope –, , , –, , , –, –, , , –, –,  ambiguity , –, –, – commutativity  economy ,  of negation , –, , , , , –, , –, –, , , , –, , , , –, , –, , , , , , –, , , ,  of affixal negation  reconstruction , –,  wide –, , , ,  scoreboard discourse model second position ,  semantic scope –,  semantics –,  of affixal negation –, ,  semimodal  Sentani  sentence-picture verification task , –, ,  sentences , ,  affirmative –, , , –, , , , –,  negative , , , –, , –, –, – sentential connective  sententialism – senza – Serbian/Croatian , ,  Sicilian , , –, ,  Σ , –, , –,  sign languages , – simplicity , ,  simulation – single-pair interpretation – Sino-Tibetan ,  situation semantics , , – situations , , – Slavic , , , , –,  “Slingshot” argument see Godel’s Slingshot social reasoning , –, – Spanish , , , , , –, –, –, , , –, –, , –, , , , , , , –, , –, – Basque Country  Rioplatense –, 

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi





spatiotemporal  frame adverbial , , , ,  frames –,  region –, , , –, – speaker’s attitude ,  spectrum of negation – speech act –, , –, , –, , – split-scope –, – Squamish  square of oppositions –, , , – St’at’imcets  Standard Average European , ,  standard varieties –, –,  standardization – stativizer operator  stigmatized  still – strengthening effect –, –, – stripping, negation in  subject ,  agreement  preposing , –, – raising –, –,  raising verb  subjunctive , , , , ,  subordinate clause , –, –, , , – suffixation , – superiority – superlative  suppression –, – surface scope –, – Surmic languages , –,  swear words –, ,  symmetry , –, , ,  syncretism –, – syntactic islands , – syntactic movement –, , –, , – syntactic scope –, – syntactic tests ,  either/too-test –, ,  neither-test –, , , ,  question tag-test –, –,  not even-test –,  why not?-test  syntactic variants , , , – syntax , , , 

T

Tacana  Tag – constant – falling – reverse –, – rising –

Tahitian – Tamil  Tennet  tense –, , , ,  Tense/Aspect/Modality (TAM)  thematic relation , , ,  separation  Tirmanga  Tneg-marker –,  Tokelauan  tone , , , , , –, , ,  Tongan –, , ,  too  topic –, –, –, – transcranial magnetic stimulation –, – transitive permissive imperative  trivalent logic – trigger –, – truth-value gap – Tsaukambo ,  Tupi Guarani  Turkana  Turkish , , , , , – Turkic  Tuyuca ,  two-step model –, –

U

underspecification – uniqueness , –,  universal item  until –, , – Uralic languages –, – utterance time 

V

variability , , , –,  variable binding , – variationist sociolinguistics ,  vehicle change – Venetian ,  verb –, , –, –, – movement –, , , –, , , – placement  veridicality –, –, –, , ,  verification , , –, – verum , – Vietnamese 

W

Wanyi –, ,  Waorani  Warlpiri  weinig 

OUP CORRECTED PROOF – FINAL, 4/3/2020, SPi

 Welsh , –, ,  Middle , ,  West Flemish ,  whdependency – element , , – in-situ –, ,  interrogative –, –, –, –,  movement , – multiple wh- see question, multiple whphrases , , , , –, , , , , –

without , , , ,  word order , , –, , , , –, 

Y

yet , –

Z

Zazaki  zero quantifier  Zulu 



OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

OXFORD HANDBOOKS IN LINGUISTICS THE OXFORD HANDBOOK OF AFRICAN AMERICAN LANGUAGE Edited by Sonja Lanehart

THE OXFORD HANDBOOK OF AFRICAN LANGUAGES Edited by Rainer Vossen and Gerrit J. Dimmendaal

THE OXFORD HANDBOOK OF APPLIED LINGUISTICS Second edition Edited by Robert B. Kaplan

THE OXFORD HANDBOOK OF ARABIC LINGUISTICS Edited by Jonathan Owens

THE OXFORD HANDBOOK OF CASE Edited by Andrej Malchukov and Andrew Spencer

THE OXFORD HANDBOOK OF CHINESE LINGUISTICS Edited by William S-Y Wang and Chaofen Sun

THE OXFORD HANDBOOK OF COGNITIVE LINGUISTICS Edited by Dirk Geeraerts and Hubert Cuyckens

THE OXFORD HANDBOOK OF COMPARATIVE SYNTAX Edited by Gugliemo Cinque and Richard S. Kayne

THE OXFORD HANDBOOK OF COMPOSITIONALITY Edited by Markus Werning, Wolfram Hinzen, and Edouard Machery

THE OXFORD HANDBOOK OF COMPOUNDING Edited by Rochelle Lieber and Pavol Štekauer

THE OXFORD HANDBOOK OF COMPUTATIONAL LINGUISTICS Edited by Ruslan Mitkov

THE OXFORD HANDBOOK OF CONSTRUCTION GRAMMAR Edited by Thomas Hoffman and Graeme Trousdale

THE OXFORD HANDBOOK OF CORPUS PHONOLOGY Edited by Jacques Durand, Ulrike Gut, and Gjert Kristoffersen

THE OXFORD HANDBOOK OF DERIVATIONAL MORPHOLOGY Edited by Rochelle Lieber and Pavol Štekauer

THE OXFORD HANDBOOK OF DEVELOPMENTAL LINGUISTICS Edited by Jeffrey Lidz, William Snyder, and Joe Pater

THE OXFORD HANDBOOK OF ENDANGERED LANGUAGES Edited by Kenneth L. Rehg and Lyle Campbell

THE OXFORD HANDBOOK OF ELLIPSIS Edited by Jeroen van Craenenbroeck and Tanja Temmerman

THE OXFORD HANDBOOK OF ENGLISH GRAMMAR Edited by Bas Aarts, Jill Bowie, and Gergana Popova

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

THE OXFORD HANDBOOK OF ERGATIVITY Edited by Jessica Coon, Diane Massam, and Lisa deMena Travis

THE OXFORD HANDBOOK OF EVENT STRUCTURE Edited by Robert Truswell

THE OXFORD HANDBOOK OF EVIDENTIALITY Edited by Alexandra Y. Aikhenvald

THE OXFORD HANDBOOK OF EXPERIMENTAL SEMANTICS AND PRAGMATICS Edited by Chris Cummins and Napoleon Katsos

THE OXFORD HANDBOOK OF GRAMMATICALIZATION Edited by Heiko Narrog and Bernd Heine

THE OXFORD HANDBOOK OF HISTORICAL PHONOLOGY Edited by Patrick Honeybone and Joseph Salmons

THE OXFORD HANDBOOK OF THE HISTORY OF ENGLISH Edited by Terttu Nevalainen and Elizabeth Closs Traugott

THE OXFORD HANDBOOK OF THE HISTORY OF LINGUISTICS Edited by Keith Allan

THE OXFORD HANDBOOK OF INFLECTION Edited by Matthew Baerman

THE OXFORD HANDBOOK OF INFORMATION STRUCTURE Edited by Caroline Féry and Shinichiro Ishihara

THE OXFORD HANDBOOK OF JAPANESE LINGUISTICS Edited by Shigeru Miyagawa and Mamoru Saito

THE OXFORD HANDBOOK OF LABORATORY PHONOLOGY Edited by Abigail C. Cohn, Cécile Fougeron, and Marie Hoffman

THE OXFORD HANDBOOK OF LANGUAGE AND LAW Edited by Peter Tiersma and Lawrence M. Solan

THE OXFORD HANDBOOK OF LANGUAGE AND SOCIETY Edited by Ofelia García, Nelson Flores, and Massimiliano Spotti

THE OXFORD HANDBOOK OF LANGUAGE ATTRITION Edited by Monika S. Schmid and Barbara Köpke

THE OXFORD HANDBOOK OF LANGUAGE CONTACT Edited by Anthony P. Grant

THE OXFORD HANDBOOK OF LANGUAGE EVOLUTION Edited by Maggie Tallerman and Kathleen Gibson

THE OXFORD HANDBOOK OF LANGUAGE POLICY AND PLANNING Edited by James W. Tollefson and Miguel Pérez-Milans

THE OXFORD HANDBOOK OF LEXICOGRAPHY Edited by Philip Durkin

THE OXFORD HANDBOOK OF LINGUISTIC ANALYSIS Second edition Edited by Bernd Heine and Heiko Narrog

OUP CORRECTED PROOF – FINAL, 3/3/2020, SPi

THE OXFORD HANDBOOK OF LINGUISTIC FIELDWORK Edited by Nicholas Thieberger

THE OXFORD HANDBOOK OF LINGUISTIC INTERFACES Edited by Gillian Ramchand and Charles Reiss

THE OXFORD HANDBOOK OF LINGUISTIC MINIMALISM Edited by Cedric Boeckx

THE OXFORD HANDBOOK OF LINGUISTIC TYPOLOGY Edited by Jae Jung Song

THE OXFORD HANDBOOK OF LYING Edited by Jörg Meibauer

THE OXFORD HANDBOOK OF MODALITY AND MOOD Edited by Jan Nuyts and Johan van der Auwera

THE OXFORD HANDBOOK OF MORPHOLOGICAL THEORY Edited by Jenny Audring and Francesca Masini

THE OXFORD HANDBOOK OF NAMES AND NAMING Edited by Carole Hough

THE OXFORD HANDBOOK OF PERSIAN LINGUISTICS Edited by Anousha Sedighi and Pouneh Shabani-Jadidi

THE OXFORD HANDBOOK OF POLYSYNTHESIS Edited by Michael Fortescue, Marianne Mithun, and Nicholas Evans

THE OXFORD HANDBOOK OF PRAGMATICS Edited by Yan Huang

THE OXFORD HANDBOOK OF REFERENCE Edited by Jeanette Gundel and Barbara Abbott

THE OXFORD HANDBOOK OF SOCIOLINGUISTICS Second Edition Edited by Robert Bayley, Richard Cameron, and Ceil Lucas

THE OXFORD HANDBOOK OF TABOO WORDS AND LANGUAGE Edited by Keith Allan

THE OXFORD HANDBOOK OF TENSE AND ASPECT Edited by Robert I. Binnick

THE OXFORD HANDBOOK OF THE WORD Edited by John R. Taylor

THE OXFORD HANDBOOK OF TRANSLATION STUDIES Edited by Kirsten Malmkjaer and Kevin Windle

THE OXFORD HANDBOOK OF UNIVERSAL GRAMMAR Edited by Ian Roberts

THE OXFORD HANDBOOK OF WORLD ENGLISHES Edited by Markku Filppula, Juhani Klemola, and Devyani Sharma